The Caddoan, Iroquoian, and Siouan Languages 9783110804669, 9789027934437

Citation preview

Trends in Linguistics State-of- the-Art Reports edited by

W Winter University of Kiel, Germany

3

THE CADDOAN, IROQUOIAN, AND SIOUAN LANGUAGES

by

WALLACE L. CHAFE

1976 MOUTON THE H A G U E - P A R I S

© Copyright 1976 Mouton & Co. Β. V., Publishers, The Hague No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publishers

ISBN 90 279 3443 6

Printed in The Netherlands

TABLE OF CONTENTS

INTRODUCTION 1 THE CURRENT STATE OF RESEARCH 1.1 The state of Caddoan research 1.1.0 Introduction 1.1.1 Caddo 1.1.2 Wichita 1.1.3 Kitsai 1.1.4 Pawnee 1.1.5 Arikara 1.1.6 Relationships within Caddoan 1.2 The state of Iroquoian research 1.2.0 Introduction 1.2.1 Laurentian 1.2.2 Huron 1.2.3 Mohawk 1.2.4 Oneida 1.2.5 Onondaga 1.2.6 Cayuga 1.2.7 Seneca 1.2.8 Tuscarora 1.2.9 Northern Iroquoian in general 1.2.10 Cherokee 1.2.11 Relationships within Iroquoian 1.3 The state of Siouan research 1.3.0 Introduction 1.3.1 Dakota 1.3.2 Chiwere (Iowa, Oto, and Missouri) 1.3.3. Winnebago 1.3.4 Dhegiha 1.3.5 Crow

7 11 11 11 11 13 13 14 15 15 16 16 17 18 19 20 21 22 23 24 24 25 27 28 28 29 32 33 34 35

6 1.3.6 1.3.7 1.3.8 1.3.9 1.3.10 1.3.11

Hidatsa Mandan O f o and Biloxi Tutelo Catawba Relationships within Siouan

2 REMOTE RELATIONSHIPS 2.1 The Macro-Siouan hypothesis 2.1.0 Introduction 2.1.1 Siouan and Iroquoian 2.1.2 Siouan and C a d d o a n 2.1.3 Iroquoian and C a d d o a n 2.1.4 Macro-Siouan phonology 2.2 Other possible remote relationships

36 37 38 39 39 40 43 43 43 43 44 47 52 53

3 A BRIEF LOOK AT THE CADDO LANGUAGE 3.1 C a d d o phonology 3.1.0 General remarks 3.1.1 Surface and underlying phonological features 3.1.2 Significant phonological processes 3.2 C a d d o grammar 3.2.0 General remarks 3.2.1 Person, case, and reality 3.2.2 N u m b e r 3.2.3 Incorporation 3.2.4 Benefactive 3.2.5 Sitting, standing, and lying 3.2.6 Tense and aspect 3.2.7 Lexical gap questions 3.2.8 Nominalization 3.2.9 Adverbial clauses 3.2.10 Validity 3.2.11 Summary

55 55 55 56 58 64 64 65 68 70 72 74 75 77 7g 80 81 82

4 REFERENCES Addenda

83 96

INTRODUCTION

If one looks at the earliest recorded distribution of Indian languages in the United States and southern Canada and subtracts the far-flung Algonkian languages and the Muskogean languages of the Southeast, the remaining territory east of the Rocky Mountains is found to be covered principally by three major language families. The area of the Great Plains contains two of these families, Siouan and Caddoan, overlapping to some extent, but with the Siouan languages for the most part to the north of the Caddoan. Both Siouan and Caddoan are represented also in the Eastern Woodlands area. The third family, Iroquoian, is entirely a woodlands family, with a center of gravity in the Northeast but a single important representative, Cherokee, farther to the south. This volume deals with these three language families. It begins in Part 1 with a discussion of the research that has so far been done on the languages included in each of them. I have concentrated on research that has been reported in the published literature, although some particularly important unpublished material and guides to such material are mentioned, as are those current research projects of which I have been aware. I have taken into account not only descriptive work but also whatever comparative work has been done, as well as work dealing with subgrouping within each of these families. In Part 2 I discuss various remote relationships which have been hypothesized to exist between each of these three families and other language families. I give most attention to the possibility that the Caddoan, Iroquoian, and Siouan families are remotely related to each other within a larger stock that might be called Macro-Siouan. It seems to me that the evidence for such a larger grouping, while certainly lacking in quantity, is nevertheless of such a quality as to be compelling. It is thus likely that Sapir was right in positing a tie between Iroquoian and Caddoan (Sapir 1951: 173), but I suggest that there is at least as much reason to believe that Siouan is a part of this same picture. In Chapter 2.2 I mention work that might suggest ties between Macro-

8 Siouan and still other language families. It is the Macro-Siouan hypothesis that gives this volume its unity. When one studies different languages from two or all three of these families, one does indeed gain the impression that their structures have much in common. Presumably it was this kind of feeling that led Sapir to group together Iroquoian and Caddoan. It is impossible within the scope of this volume to give any kind of complete picture of what the structure of these languages is like. There is of course a great deal of diversity among them, even within the same family, and any comprehensive presentation of similarities and differences would require a major work of large proportions. What I have chosen to do instead is to present in Part 3 a brief sketch of a few of the more interesting features of only one language, Caddo, as an illustration of the kind of structure these languages exhibit. Various of the features described for Caddo are fairly typical of MacroSiouan languages. I might mention above all their polysynthetic and fusional character (Sapir 1921: 120-46). With respect to polysynthesis, the surface-structure verbs of these languages have absorbed a variety of elements which belong semantically outside the verb. Thus, these verbs are likely to include various features belonging semantically to associated nouns. Such features are likely to be evident especially in verbal prefixes which express the person, number, case, and sometimes gender of one or more associated nouns. The number of such prefixes is smallest in Siouan, somewhat larger in Caddoan, and reaches its height in the Northern Iroquoian languages where approximately 65 different prefixes of this kind are to be found. The tendency of surface structure verbs to attract elements which are semantically outside the verb has its extreme manifestation in so-called noun incorporation, whereby even the lexical content of certain nouns is absorbed into the verb. This tendency too is particularly evident in the Northern Iroquoian languages, but we will see some manifestation of it in Caddo. The verbs of these languages are also likely to contain various elements expressing tense, aspect, modality, subordination, and the like, and again Caddo provides us with some examples. With respect to fusion, it has been the tendency of the rather long words which result from polysynthesis to undergo sound change, so that their constituent elements are 'fused' together. Since both the degree and the specific kind of fusion is dependent on the amount and kind of sound change, even closely related languages may show considerable differences as far as specifics are concerned. The tendency toward fusion of some kind is, however, widespread, and Caddo provides excellent illustrations. Of course neither polysynthesis nor fusion is unique to the Macro-

9 Siouan languages. The bonds that unite these languages are more particular, as illustrated in Part 2. But at least these characteristics are typical of many languages in this family, and the discussion of Caddo will provide some general insight into their nature. I have chosen Caddo partly because of the excellent way in which it illustrates both polysynthesis and fusion, but partly also because there exists at the present time virtually no published material on this language. The reader will be able to go to a variety of sources, as described in Part 1, for information on Iroquoian and Siouan languages, but he will have much less to refer to so far as the Caddoan languages are concerned, and essentially nothing for Caddo itself. The sketch in Part 3, then, will serve also to document an almost unknown language in a few of its more salient aspects. Parts 1 and 2 of this work appear in a slightly different form in Chafe (1973). I am grateful to Floyd G. Lounsbury and Allan R. Taylor for their suggestions of material that has been added to Chapters 1.2 and 1.3.

THE CURRENT STATE OF RESEARCH

THE STATE O F C A D D O A N RESEARCH 1.1.0

Introduction

There are essentially three extant Caddoan languages : Caddo, Wichita, and Pawnee. A geographically separated and quite aberrant dialect of Pawnee known as Arikara (or Ree) is usually listed as a separate language. Another language, Kitsai, lost its last speaker a few decades ago. Caddo, Wichita, and Pawnee are now spoken in Oklahoma, Arikara in North Dakota. There are no more than a few hundred speakers of any of these languages remaining ; in the case of Wichita and Arikara the number is smaller than that. None of the languages is being learned by children. An excellent discussion of all but the most recent history of Caddoan linguistics is Taylor (1963a), on which much of the following is based.

1.1.1

Caddo

The ancestors of the present Caddo, when first encountered by Europeans, formed a far-flung group of individual bands of varying size on the western edge of the Southeast culture area. Aside from a few isolated groups, these bands were clustered into several larger entities, chief among which were the Hasinai, in what is presently eastern Texas, and the Kadohadacho, farther to the north in approximately the area where the Red River now forms the boundary between Texas and Arkansas. Other bands lived farther south on the Red River in Louisiana and farther east and north in Arkansas. Linguistically, these bands differed at least to the extent that they spoke a number of distinct dialects. The present Caddo represent a melting pot of these different bands, created by an amalgamation of all of them in Texas in the nineteenth century. In 1859 they fled from Texas into what is now Oklahoma, where they were

12 settled in the neighborhood of the present towns of Anadarko, Binger, and Ft. Cobb. Remnants of the earlier dialect differences are still found distributed among them in a manner that would repay further study. A group of Caddo may have been visited by the DeSoto expedition as early as 1541, but these people seem not to have been seen by Europeans again until the 17th century, when a variety of contacts took place between them and both Spanish and French explorers and traders. It seems that LaSalle may have been the first to record some words of their language, in 1687. It was not until the beginning of the 19th century, however, that we have anything more than a few sporadic words preserved. During the first decade of that century, just after the Louisiana Purchase, John Sibley collected for Thomas Jefferson vocabulary lists of Caddo, of a divergent dialect called Natchitoches, and of a distinct language called Adai whose affiliation remains problematic. The Natchitoches vocabulary has been lost, but the Caddo list was later published (Claiborne - Mason 1879), as was the Adai (Gallatin 1836: 307-67). Sibley was the first Indian agent in this area. One of his successors, George Gray, also collected a Caddo vocabulary and some sentences, and it was this material (with nineteen additional words from Sibley) which was published by Gallatin (1836: 307-67, 383-97, 409-13). Schoolcraft (1853: 709-12) published a Caddo vocabulary which had been collected by Captain Randolph B. Marcy. Twenty words of Caddo collected by Lieutenant Amiel Weeks Whipple were also published at about the same time (Whipple 1856: 70). Still later in the 19th century, James Mooney included a short Caddo glossary in his ghost-dance study (Mooney 1896: 1102-3). In the present century, a set of Caddo kinship terms was discussed by Leslie Spier ( 1924), and a number of Caddo words were included among some ethnographic notes published by Elsie Clews Parsons (1941). In 1956 and early 1957 Daniel da Cruz, at that time an undergraduate student at Georgetown University, worked with a Caddo informant and subsequently wrote a provisional phonemic analysis of the language as a senior essay (da Cruz 1957). In the first half of the 1960's Wallace L. Chafe spent four summers working with the language. The results of that work should eventually be published, but at present reference can be made only to some Caddo examples in a work dealing with phonological theory (Chafe 1968). See also Part 3 of this volume.

13 1.1.2

Wichita

The Wichita are generally thought to have been the inhabitants of the 'Province of Quivira', which represented the farthest penetration into the North American continent by the Coronado expedition in 1541. At that time the Wichita were probably located in central Kansas. After moving into what is now Oklahoma, where they were found by Frenchmen in the early 18th century, they were gradually forced further south into Texas before the end of that century. Later they returned to Oklahoma, fled to Kansas during the Civil War, and were finally placed together with the Caddo in the vicinity of Anadarko, Oklahoma, particularly in and about the town of Gracemont. As with the Caddo language, there is a small amount of poorly transcribed Wichita vocabulary material from the 19th century. In particular, there are two different vocabulary lists collected by Marcy (Marcy 1853: 307-8; Schoolcraft 1953: 709-11), plus Whipple's vocabulary of the now extinct dialect called Waco (Whipple 1856: 65-8). Spier's article on kinship (Spier 1924) dealt with Wichita as well as Caddo terms. Edward S. Curtis (1907-30 19: 230-7) also gives a Wichita vocabulary. In 1949 the Wichita were visited by Paul L. Garvin, who subsequently published an article on Wichita phonemes (Garvin 1950), and who also recorded the kinship terms that were discussed in Schmitt (1952). More recent and more extensive work has been done by David S. Rood, who began field work on the language in 1965, and who has completed a grammar of the language as a doctoral dissertation (Rood 1969).

1.1.3

Kitsai

The Kitsai may have been located prehistorically in Oklahoma, but Europeans first found them living in what is now Texas between the Red River and the upper Trinity. In 1858 they fled to Oklahoma, where they have been living among the Wichita ever since. In recent times, at least, the Wichita have considered the Kitsai to be part of their own tribe, and all the last speakers of Kitsai also spoke Wichita. The last fluent speaker of Kitsai evidently died during the 1930's. A vocabulary of Kitsai appears in Whipple (1856: 65-8). Kitsai data were recorded by Alexander Lesser in the summers of 1929 and 1930. His extensive notes, to be deposited in the Library of the American Philosophical Society,

14 provide the only important source of information on this language. Bucca - Lesser (1969) is based on this material.

1.1.4

Pawnee

Coronado's guide through the plains area in the 16th century is thought to have been a Pawnee. Significant European contact, however, awaited the coming of French explorers in the early 18th century. At that time the Pawnee were located principally in the area of what is now Nebraska, in the neighborhood of the Platte River. They seem to have been divided into four distinct bands (or five, including the Arikara, to be discussed separately below). Three of these bands spoke approximately the same dialect. The fourth, known as the Skidi (Skiri), spoke a distinguishable but not radically different dialect. In the 19th century the Pawnee moved to what is now Oklahoma, where they presently live in and about the town of Pawnee. The first systematic recording of the Pawnee language seems to have been a list of vocabulary items collected for Thomas Jefferson by the Lewis and Clark expedition in 1804-05. The earliest published vocabulary was collected by Thomas Say, a member of the Long expedition in 18191820, and was published in the report on that expedition by Edwin James (Thwaites 1904-07 17: 290-8, 305). Say's vocabulary also appeared in Gallatin (1836: 307-67). The German nobleman, Alexander Philipp Maximilian, Prinz von Wied, during his travels in the West in the 1830's, collected a brief Pawnee vocabulary from a non-Indian who spoke Pawnee (Wied 1839-41 2: 630-2; also in Thwaites 1904-07 24 : 293-4). A lengthier vocabulary of Pawnee was published by Ferdinand Vandeveer Hayden (186ß: 347-51), who received his material from William Hamilton, a missionary among the Pawnee. Hayden himself collected additional Pawnee material, both lexical and grammatical, which he published a few years later (Hayden 1869: 390-406). Lewis Henry Morgan (1871: 293-382) included kinship terms from two different bands of Pawnee collected by himself and B. F. Lushbaugh. A short grammar of Pawnee by John Brown Dunbar appeared in 1890 as an appendix to a collection of Pawnee stories (Grinnell 1893: 409-37). Texts of some Pawnee songs are found in Fletcher (1902) and Densmore (1929). Gilmore (1919: 139-45, 149-50) gives the Pawnee names of some plants. Gene Weltfish has devoted much of her career to ethnographic (including linguistic) study of the Pawnee. She published a Pawnee text with a detailed grammatical analysis (Weltfish, 1936), and some further texts with

15 word for word translations (Weltfish 1937). Douglas R. Parks began working with the language in the summer of 1965, and is, at the time of this writing, preparing a description of it as a doctoral dissertation.

1.1.5

Arikara

Beginning as a northern offshoot of the Pawnee, the Arikara have in historic times gradually moved northward along the Missouri River from South Dakota into North Dakota. They moved to Fort Berthold in North Dakota in 1862, and eventually became one of the three tribes (with the Mandan and Hidatsa) on the Fort Berthold Reservation. Gallatin mentioned the Arikara, saying, All the accounts of the Indians and of the interpreters agree in the fact of their speaking Pawnee, but we have no vocabulary of their language. (Gallatin 1836: 129.)

A significant vocabulary had, however, been collected by Wied shortly before Gallatin's work was published. Unusually accurate for the time, it was first published in Wied (1839^1 : 465-74), and can be found also in Thwaites (1904-07 24 : 210-4). Another list of Arikara words, less satisfactorily recorded than those of the Wied list, was published by the well-known painter of Indian subjects, George Catlin (1841 2: 262-5). There is evidence that this list was made, not by Catlin, but by an official of the American Fur Company named Kenneth Mackenzie. (Latham 1846: 32.) A third Arikara vocabulary appeared in Hayden (1863: 35663). Morgan (1871: 293-382) includes Arikara kinship terms. Curtis (1907-30 5: 169-77) also contains an Arikara vocabulary. In more recent years occasional brief field work on Arikara has been done by Gene Weltfish, Melvin Gilmore, Allan R. Taylor, Wallace L. Chafe, and Douglas R. Parks. Parks contemplates further work with Arikara as an extension of his Pawnee studies.

1.1.6

Relationships within Caddoan

The relationship of Wichita and Kitsai to Pawnee is not so distant that it would escape any scholar in possession of sufficient data from these three languages. The first published mention of this relationship seems to have been that included in a footnote in Josiah Gregg's Commerce of the Prairies :

16 The Pawnees and Rickaras of the north, and the Wacoes, Wichitas, Towockanoes, Towyash and Keechyes, of Red River, are of the same origin. (Gregg 1844 2: 251.)

Documentation was provided in Whipple - Ewbank - Turner (1855: 68-9), based on the vocabulary collected by Whipple. In that same work, Turner also mentioned the possible inclusion of Caddo in this family, and he gave a list of half a dozen possible cognate sets (WhippleEwbank - Turner 1855:70). Subsequent classifications of Indian languages accepted the inclusion of Caddo, at first tentatively, but as time went on with increasing conviction (Latham 1860: 400; Buschmann 1859: 448; Keane 1878: 478; Gatschet 1884: 42; Brinton 1891: 95-7). During this period reference was usually made to the Pawnee family or stock, but the Powell classification established the name Caddoan, which has been used ever since (Powell 1891: 58-62). The inclusion of Adai in the Caddoan family has been at various times suggested. Our knowledge of the language is so limited, however, that no definitive statement on this point may ever be possible (see discussion in Taylor 1963a: 57-8). As far as subgrouping within the Caddoan family is concerned, the statement of Lesser and Weltfish is probably as accurate as any that can be given at the present time : Pawnee, Wichita, and Kitsai are, in relation to each other, about equally divergent, save that Kitsai in phonetic structure and some forms is probably closer to Pawnee than Wichita is to Pawnee. All three, however, are mutually unintelligible. Caddo is the most divergent of the four languages. (Lesser - Weltfish 1932: 1.)

The only systematic comparative work on the Caddoan languages is Taylor (1963b), which does not take into account the more recent descriptive work of Chafe, Parks, and Rood.

1.2 1.2.0

THE STATE O F IROQUOIAN RESEARCH

Introduction

The Iroquoian languages divide themselves into two major branches, Northern Iroquoian and Southern Iroquoian. The degree of difference between these two branches is only a little less than that between the most divergent languages of the Indo-European family. The Northern Iroquoian languages which are still spoken are six in number. They include the very closely related languages of the original Five Nations

17

of the Iroquois — Mohawk, Oneida, Onondaga, Cayuga, and Seneca — plus Tuscarora, now the Sixth Nation, a language somewhat more divergent. These languages are spoken largely in New York State and neighboring Canada, although there are a number of Oneida speakers in Wisconsin and a few speakers of Cayuga in Oklahoma. None of the languages has more than a few thousand speakers, most of them middleaged or older, and Onondaga, Cayuga, and Tuscarora have well under a thousand, with Tuscarora probably the closest to extinction. A seventh Northern Iroquoian language, Huron (Wyandot), was spoken by a few individuals in Oklahoma until recently. In earlier times the number of Northern Iroquoian languages was larger. Various other languages of this family were spoken in Ontario and Quebec, in western New York, in Pennsylvania, and northern Ohio, as well as in parts of Maryland, eastern Virginia, and eastern North Carolina. All these other languages were obliterated in the early days of European colonization, and we know next to nothing about them. A short vocabulary of one group living along the Susquehanna River, variously called Susquehannock, Conestoga, Andaste, or Minqua, is preserved in Campanius (1696: 155— 60). Gallatin (1836: 305-67) gives some words from Nottoway, a language spoken in southeastern Virginia and apparently closely related to Tuscarora. The record of another language on the St. Lawrence River is discussed immediately below under 'Laurentian' (1.2.1). There is only one Southern Iroquoian language, Cherokee, although it is spoken in at least half a dozen dialects. There are roughly ten thousand speakers of Cherokee of all ages, most of them in eastern Oklahoma, but some still occupying a small part of their original territory in western North Carolina.

1.2.1

Laurentian

The first North American language to be recorded by a European was a Northern Iroquoian language. Jacques Cartier, after his voyage to the Gulf of St. Lawrence in 1534, took back to France two young men who were part of a fishing party which was temporarily in that area, but whose home was farther to the west. After his second voyage in 1535-36 he returned to France with other captives from the neighborhood of Quebec City. The accounts of these two voyages have appended to them vocabularies which may have been elicited in France from some of these people by the first North American Indian linguist, whoever he may have been. They are published in Biggar (1924: 80-1, 241-6).

18 The language of these 'Carter vocabularies' is often assumed to have been Huron, or partly Huron and partly Mohawk (Barbeau 1959), but it seems more likely that it was some third language, not unlike Huron, that was spoken along the St. Lawrence River in the early 16th century (Chafe 1962; Tooker 1964: 3 ^ ) . Champlain, at the beginning of the 17th century, did not find Iroquoian speakers in this area. The name Laurentian has been used to refer to this earliest recorded Iroquoian language. During the 17th, 18th, and 19th centuries a great deal of missionary work was performed among the speakers of Northern Iroquoian languages. The missionaries focused much attention on linguistic matters, for they were generally desirous of preaching to the Indians in their own language, in translating sections of the Bible, prayers, and hymns, and in providing language materials for the training of new missionaries. A number of grammars, vocabularies, and religious texts of various sorts remain in manuscript. The interested reader should consult both Pilling (1888) and Hanzeli (1969) for lists and discussions of these and other sources, both published and unpublished. The summaries below will focus on those published sources which are most directly relevant to linguistic studies. Adelung - Vater (1806-17) will not be cited for each language, but a listing of its various Iroquoian vocabularies, reprinted from other sources, is given in Pilling (1888: 2).

1.2.2

Huron

The Huron, on first European contact, occupied an area in present Ontario southeast of Georgian Bay. They were visited by Champlain in 1615, but more important linguistically was the visit of the Recollet missionary Gabriel Sagard-Théodat in 1623-24. Sagard prepared a dictionary of the Huron language that was first published a few years later (Sagard 1632). In 1634 Jesuit missionaries from France began intensive work among the Huron, which lasted until the devastating defeat of these people by the Iroquois in 1649-50. Many Huron words are scattered through the Jesuit Relations for this period (Thwaites 1896-1901). Of particular interest are the comments on the language by Jean de Brébeuf (Thwaites 1896-1901 10: 117-23; also in Gallatin 1836: 236-8), and the grammar prepared by Pierre Joseph Marie Chaumonot, which was evidently the source of a work published in translation at a much later date (Wilkie 1831). The Huron who survived the defeat at the hands of the Iroquois

19

scattered in all directions, and groups of them eventually settled in a variety of locations. In the mid-18th century another Jesuit, Pierre Potier, working at the Huron mission at Sandwich, Ontario, put together an extensive grammar and dictionary, based to a large extent on earlier manuscripts, especially Chaumonot's. A facsimile edition of Potier's manuscripts is now available (Fraser 1920). Potier's work can be viewed as the culmination of all the Jesuits' linguistic work among the Huron. Another Huron group, after many vicissitudes, finally settled in northeastern Oklahoma, where they have been known as the Wyandot. The language spoken there was recorded in 1911-12 by Marius Barbeau of the National Museum of Canada. This material was used in Barbeau (1915), a work comparing Wyandot 'pronominal prefixes' with those in some other Northern Iroquoian languages. Texts collected at that time were published in Barbeau (1960). Barbeau's lexical material remains in manuscript form in the National Museum of Canada.

1.2.3

Mohawk

In earliest recorded times the Mohawk, the easternmost of the Five Nations, occupied a half dozen or more villages in the Mohawk River valley in present New York State, approximately from Schenectady to Utica. They were first encountered by Champlain in 1609, and shortly after by the Dutch. Provided with guns by the Dutch and later the English, they occupied an important military position until the time of the American Revolution. Having sided with the English at that time (as did most of the other Iroquois, except for the Oneida and Tuscarora), many of them then moved to Canada, where they lived with other Iroquois on the Grand River Reserve in Ontario. Still in the 17th century a group of converted Mohawk were settled by the Jesuits at Caughnawaga near Montreal. In the mid-18th century an offshoot from this group formed a settlement further up the St. Lawrence, and their descendants now inhabit the St. Regis Reservation in northernmost New York and adjacent Canada. Smaller groups of Mohawk settled at Tyendinaga and Gibson in Ontario and at Oka in Quebec. Still others now live in Brooklyn, attracted by high steel work. The first significant linguistic work on Mohawk was carried on by Jacques Bruyas, a Jesuit, whose dictionary (Bruyas 1863), with a few accompanying grammatical notes, was compiled during the 1670's. In the next century a Moravian missionary, Johann Christoph Pyrlaeus, came up from Pennsylvania to work with the Mohawk, but none of his

20 extensive notes on the language have been published. In the 19th century Jean André Cuoq, a Sulpician missionary, worked at Oka on both the Mohawk and Algonkin languages. Several scholarly works by Cuoq on the nature of Indian languages drew heavily on both these languages for examples. From the point of view of Mohawk descriptive material, Cuoq's most important publications are his chapters on Mohawk grammar (Cuoq 1866) and his Mohawk dictionary (Cuoq 1882). Cuoq's work was based in part on manuscripts left by Joseph Marcoux, a missionary at St. Regis and Caughnawaga earlier in the century. Other 19th century sources on Mohawk include the vocabularies published by Gallatin (1836: 305-67, 383-97) and by Schoolcraft (1846: 264- 70; 1851-57 2: 482-93), the kinship terms in Morgan (1871: 291-382), and Horatio Hale's grammatical discussion and glossary (Hale 1883b: 99113, 191-215). Numerous minor sources are listed in Pilling (1888: 121-3). In the present century we have, first of all, the long and carefully transcribed cosmological texts published by J. N. B. Hewitt, one of which is in Mohawk (Hewitt 1903: 255-339), plus a short ritual text (Hewitt 1928b). Huot (1948) discusses the manner in which Mohawk words have been adapted or invented to apply to items of white culture. Lounsbury (1960) gives an excellent discussion of Mohawk place-names in the Champlain Valley. Considerable work on Mohawk has been done by Paul M. Postal. His dissertation (Postal 1962) remains unpublished. There have appeared in print an article on the generation of Mohawk prefixes (Postal 1964b) and on vowel doubling (Postal 1969), as well as discussions of various points of Mohawk phonology in an article on Boas's phonological practice (Postal 1964a) and in a book on phonological theory (Postal 1968). Very recently a Mohawk grammar was completed as a dissertation by Nancy Bonvillain (1972). Both she and John Beatty are conducting further work on the language.

1.2.4

Oneida

Earliest records show the Oneida settled south of Oneida Lake in New York. A remnant of them remains in their original location in Oneida County, New York. Others, however, joined the Iroquois on the Grand River Reserve in Ontario, and still another group entered Canada in 1849 to settle on the Thames River, where they are known as the Oneidas of the Thames. The largest number of Oneida, however, left

21 New York in 1846 and travelled to land on Green Bay, Wisconsin, where they remain today. The early missionary grammars and dictionaries that we have for Mohawk are lacking for Oneida, which is, however, so closely related that it verges on being an aberrant Mohawk dialect. Nineteenth century vocabularies include those published in Gallatin (1836: 305-67), Schoolcraft (1846: 279-81), and Schoolcraft (1851-57 2: 482-93), as well as the kinship terms in Morgan (1871: 291-382). Further sources are listed in Pilling (1888: 132). In the present century we have, first of all, a grammatical sketch by Franz Boas (1909), and the authoritative description of the Oneida verb by Floyd G. Lounsbury (1953). This last work established much of the framework and terminology followed by subsequent Iroquoian scholars.

1.2.5

Onondaga

The earliest known home of the Onondaga was in the same area in New York of which they occupy a small part today. Its center was more or less the present Onondaga County, and it extended northward to Lake Ontario. After the American Revolution some of the Onondaga moved to the Grand River Reserve in Ontario, while some remained in New York, where they now live on a small reservation just south of Syracuse. What may be the earliest record of the Onondaga language is a dictionary believed to have been compiled by one of the Jesuits at the end of the 17th century. It was published by Shea (1860), who says concerning the identity of the language that it was asserted to be Onondaga by a missionary at Caughnawaga who was a competent Mohawk scholar, and who noted as the most striking differences the substitution of h for the Mohawk r, and in the preterites of i for the Mohawk on. (Shea I860: 00.)

(The h referred to is that of the beginning of the masculine pronominal prefix, and the i is that of an aspect suffix; both are indeed features by which Onondaga differs from Mohawk.) In the mid-18th century the Onondaga were visited by a Moravian missionary, David Zeisberger, whose grammar and dictionary were also published during the 19th century (Horsford 1887, Zeisberger 1888). It is a curious fact that both the dictionary Shea (1860) and the material in Zeisberger (1888) show a

22 language which differs from Onondaga as it is spoken today. Most notably, the language of the earlier records contains r, whereas in modern Onondaga, as in Seneca, r has been lost in all environments. Whether Onondaga has undergone a highly significant sound change since the mid-18th century, or whether the language described by Zeisberger (and the Shea dictionary) is not directly ancestral to modern Onondaga remains problematic. Gallatin (1836: 305-67) contains an Onondaga vocabulary taken from Zeisberger's materials, but Schoolcraft (1851-57 2: 482-93) gives material obtained independently. Again Morgan (1871: 291-382) provides a list of kinship terms. Pilling (1888: 133) can be consulted for further, less important sources. Toward the end of the 19th century a great deal of ethnographic work was done by a clergyman named William M. Beauchamp, and he published various works that contain many items of Onondaga vocabulary. An 'ethnobotanical' example is Beauchamp (1902). Hewitt published two cosmological texts in Onondaga (1903: 141-220; 1928a: 612-791). A recent treatment of Onondaga is Chafe (1970a), and the language is used for exemplificatory purposes in Chafe (1970b: 268-86). Further work is currently being done by Hanni Woodbury.

1.2.6

Cayuga

Earliest records locate the Cayuga on the shores of what is now Cayuga Lake. The majority of them moved to the Grand River Reserve in Ontario after the American Revolution. Those who remained behind were scattered among the other Iroquois tribes, but some, together with other Iroquois, moved westward into Ohio and eventually settled in northeastern Oklahoma, where a small number of them still speak the language. In Oklahoma this slightly deviant dialect of Cayuga is usually called Seneca. Published sources on Cayuga are remarkably few. Gallatin (1836: 376) lists a few words, said to have been taken from Barton. Schoolcraft (1846: 271-7; 1851-57 2: 482-93) gives longer lists. Morgan (1871: 291-382) includes Cayuga kinship terms. Speck (1949) contains some Cayuga vocabulary, particularly ceremonial terms. Floyd G. Lounsbury has worked extensively with Cayuga, but has published only some illustrations of the phonology (Lounsbury 1963: 566-7). Further work is currently being performed by Michael K. Foster.

23 1.2.7

Seneca

When first encountered the Seneca were situated between the Seneca Lake area and the Genesee River. After the neighboring Erie and Neutrals were defeated during the 17th century, the Seneca spread westward to Lake Erie. After the American Revolution some of them settled on the Grand River Reserve in Ontario, but many remained in western New York, where they now inhabit the Tonawanda, Cattaraugus, and Allegany Reservations. The small Cornplanter Grant in Pennsylvania, formerly inhabited by Seneca, was recently obliterated by the Kinzua Dam. There is little record of Seneca before the 19th century. The most important early material is found in the works of Asher Wright, a protestant missionary who worked among the Seneca, first at Buffalo Creek and then at Cattaraugus, from 1831 to 1875. Wright published a great deal of material in Seneca, some of it on his own Mission Press. Most of the material was religious in nature, for example hymnals and a translation of the Gospels, but he also published a Spelling Book (Wright 1842) and even a periodical, called in English the 'Mental Elevator', which ran to at least nineteen numbers. Wright at one time devised a unique set of letters to be used in printing Seneca, but they were never adopted. Other 19th century sources on Seneca include a vocabulary published by the Quaker Halliday Jackson (1830: 114-20), two vocabularies and some sentences in Gallatin (1836: 305-67, 383— 97, 415), a vocabulary in Schoolcraft (1846: 393-400) obtained from Ely S. Parker, and Morgan's kinship terms (1871: 291-382). Morgan's classic ethnography (Morgan 1851) also contains a number of Seneca terms, as well as a chapter dealing with the language (Morgan 1851: 394-411). Other sources are listed in Pilling (1888: 153). One of the three texts in the first part of Hewitt's Iroquoian cosmology is in Seneca (Hewitt 1903: 221-54). Curtin - Hewitt (1911: 715-43, 75690) also has texts in Seneca. Curtin (1923: 513-6) contains a glossary. A number of ethnographic works subsequent to Morgan contain scattered Seneca terms. A few representative examples of such works are Harrington (1908), Waugh (1916), Parker (1910, 1913), Fenton (1953), and ConklinSturtevant (1953). Several articles on Seneca appeared in I JAL at midcentury; one is Preston - Voegelin (1949); two, based on field work, are by Nils M. Holmer (1952-53). Holmer subsequently collected all his Seneca material in a single monograph (Holmer 1954). Wallace L. Chafe's dissertation on Seneca morphology was first published in installments in I JAL (Chafe 1960-61), and later republished together with a

24 dictionary (Chafe 1967). Some ritual texts with morphological analysis were published in Chafe (1961). Chafe (1963) was intended as a practical guide to the language, containing orthographic instructions, a sketch of Seneca word structure, and a glossary of ethnographically relevant terms. Chafe (1959a) illustrated a theoretical point with Seneca examples, and Chafe (1959b, 1964b) involved principles and examples of certain kinds of historical reconstruction. 1.2.8

Tuscarora

The early colonists found the Tuscarora located in eastern North Carolina. Difficulties with the whites led to their migration to New York State early in the 18th century, where they were subsequently accepted as the sixth nation of the Iroquois League. At the end of the 18th century most of them settled on the small reservation near Niagara Falls which they occupy today. A few, however, moved to the Grand River Reserve in Ontario. The earliest Tuscarora linguistic material of any consequence seems to be John Lawson's vocabulary from the North Carolina period (Lawson 1709: 225-30). Vocabularies published in the 19th century include those in Gallatin (1836: 305-67), Catlin (1841 2: 262-5), and Schoolcraft (1846: 251-8). Morgan (1871: 291-382) includes Tuscarora kinship terms. Other sources are listed in Pilling (1888: 162-3). J. Ν. B. Hewitt, himself a Tuscarora, prepared a large dictionary of the language, but it remains in manuscript (see Pilling 1888: 81-2). More recently there is an article by Frans M. Olbrechts in Dutch dealing with Tuscarora pronominal prefixes (Olbrechts 1929). Anthony F. C. Wallace wrote an account of linguistic materials, mainly wire recordings, collected by him in 1948 (Wallace 1949), and Wallace - Reyburn (1951) contains a text in Tuscarora. Joan Fickett, a student at the State University of New York at Buffalo, wrote an M.A. thesis on Tuscarora phonology which was subsequently published (Fickett 1967). It includes a useful background discussion of the Tuscarora and their language, as well as a brief English-Tuscarora word list. Floyd G. Lounsbury has unpublished notes from his own field work on Tuscarora, and Barbara Graymont has also worked on the language. Further work is currently being done by Marianne Williams.

1.2.9

Northern Iroquoian in general

With respect to the Northern Iroquoian languages in general, there are

25 a few brief articles from the late 19th century dealing with various points of detail. Several by Erminnie A. Smith (1883a, 1883b, 1884) are worth mentioning. There is an interesting controversy between Horatio Hale and J. Ν. B. Hewitt regarding the etymology of the word Iroquois, and various other etymological points (Hewitt 1888, 1891; Hale 1888). Another article by Hewitt on Iroquoian etymology is Hewitt (1892). All these writings show a concern for detail that was characteristic of the period in American Indian linguistic studies.

1.2.10

Cherokee

Except for the rather distant linguistic relationship, the Cherokee have little in common — culturally, historically, or geographically — with the Northern Iroquoian groups discussed above. De Soto may have encountered the Cherokee in 1540, but extensive European contact awaited the settlement of Virginia and the Carolinas. In the 17th century these people occupied the southern Appalachian region in Tennessee and North Carolina, as well as neighboring parts of Virginia, South Carolina, Georgia, and Alabama. More than most tribes, the Cherokee made a concerted effort to adapt to white ways. In spite of this effort, most of them were forced to move into what is now northeastern Oklahoma in 1838-39. A few of them hid in the mountains, and were subsequently permitted to remain on the land which forms the Qualla Reservation in North Carolina. Aside from their relatively large numbers for an American Indian linguistic group, the speakers of Cherokee are of special interest because of their adoption, in 1821, of a unique writing system devised by one of their own number, George Guess (Sequoya). The system was widely used into the first part of the 20th century. In recent time familiarity with it has declined, although most recently there has been a conscious attempt to revive it. Little systematic attention was paid to the Cherokee language before the 19th century. In 1825 Samuel A. Worcester began missionary work among the Cherokee which lasted until his death in 1859. Worcester was to the Cherokee what Asher Wright was to the Seneca. He arranged for the publication of a large amount of material in the Cherokee syllabary, including portions of the Bible, hymns, and almanacs. Worcester provided most of the material on Cherokee that is scattered throughout Gallatin (1836: 241-50, 276, 305-67, 398-404, 415-21), and the notes on Cherokee included in Schoolcraft (1851-57 2: 443-56). During this same period

26 several periodicals were published in the syllabary: the Cherokee Phoenix ( 1828-34), the Cherokee Advocate ( 1844-54,1870-1906), and the Cherokee Messenger (1844-^6). Two grammars of the language were written by scholars at about this time. The first (Pickering 1830) was never completed, but was published in unfinished form. It was based on direct work with an informant. The second (Gabelentz 1852b) was based on Pickering, on Worcester's material in Gallatin, and on copies of the Cherokee Messenger. Both these grammars (the second in translation) were republished in Krueger (1963). Thomas Say, of the Long expedition in 1819-20, collected a vocabulary of Cherokee which was published in the report of that expedition (Thwaites 1904-07 17: 290-8). Lewis Henry Morgan included Cherokee kinship terms from two different dialects, obtained from missionary sources (Morgan 1871: 291-382). Other early sources are listed in Pilling (1888: 42-4). The ethnologist James Mooney worked with the Cherokee toward the end of the 19th century, and produced several important works in which a variety of linguistic material is embedded. Mention might be made of his collection of Cherokee sacred formulas, an important literary genre in the language (Mooney 1891), and the long glossary at the end of his comprehensive volume on Cherokee history and myths (Mooney 1900: 506-48). In the present century there have been, first of all, a few brief texts in Cherokee published in IJAL (Speck 1926, Olbrechts 1931). Frans M. Olbrechts also edited further sacred formulas collected by Mooney (Mooney - Olbrechts 1932). Mary R. Haas (1948), writing on classificatory verbs in Muskogee (Creek), compared them with the same phenomenon, more highly developed, in Cherokee. A series of articles on Cherokee grammar, originating at the University of Pennsylvania, appeared in IJAL during the 40's and 50's. The first (Bender - Harris 1946) dealt with the phonemes of the North Carolina dialect. The second (Bender 1949) contained a morphemic analysis of several short texts. There followed a three-part series on Cherokee verb morphology (Reyburn 1953-54). These relatively brief and tentative articles by Reyburn constitute the only modern description of the language of any consequence. Again, Floyd G. Lounsbury possesses extensive notes from his own field work on the language. Mention should also be made of the amusing article by Archibald A. Hill ( 1952) on the alleged primitivism of Cherokee. A number of recent publications by Jack Frederick Kilpatrick and Anna Gritts Kilpatrick contain miscellaneous linguistic items (for example Kilpatrick - Kilpatrick 1965; see also Willard Walker 1967). Gulick (1958) presents a sociolinguistic study of the use of the language in North Carolina. A recent discussion of the language and the Sequoya syllabary,

27 including a grammatical sketch, is available in manuscript in Walker (1971), which also contains a useful bibliography. I will not attempt to list the various works that have dealt with the syllabary and its use, but mention can be made of the recent articles by Chafe and Kilpatrick (1963), and by John K. White (1962). A manuscript bibliography and history of Cherokee printing has been prepared by Raymond Yamachika (1961).

1.2.11

Relationships within Iroquoian

The Northern Iroquoian languages are so closely related that the fact of their relationship has always been obvious. It has long been clear, also, that the languages of the original Five Nations belong together in a subgroup as opposed, for example, to Tuscarora. The precise relations within that subgroup, however, are not entirely clear. The close affinity of Mohawk and Oneida is apparent, but whether Onondaga is to be more closely identified with them or with Seneca and Cayuga remains problematic, and perhaps it is best to regard these languages as forming a dialect continuum. One investigation of this question is reported in Hickerson - Turner - Hickerson (1952). Use was made of a technique whereby intelligibility across seven languages (including Tuscarora and Cherokee) was tested by means of tape recordings. Hoffman (1959) attempted to subgroup a larger collection of Iroquoian languages (including Huron, Laurentian, Andaste, and Nottaway) on the basis of the number of shared cognates in a limited vocabulary sample. Blin-Lagarde (1972) follows a similar procedure using a somewhat larger sample. The relation of Cherokee to the northern languages is not immediately obvious, although anyone reasonably familiar with Cherokee and one of the other languages could hardly help being aware of it. The first writer to have pointed it out seems to have been Benjamin Smith Barton (1797: xlv, lxvii). He mentions a few similarities between Cherokee words and words of the Six Nations languages. Gallatin (1836: 91-2) referred to Barton's statement and expressed hesitant agreement. A more substantive demonstration of the relationship was given by Horatio Hale (1883a: 26-8), who set forth a small but convincing collection of lexical and grammatical correspondences. Further correspondences were given by Albert S. Gatschet a few years later (Gatschet 1886). Additional evidence substantiating the relationship was compiled by J. N. B. Hewitt at about the same time, but was never published (see Pilling 1888: 81).

28

Aside from the small amount of comparative material in Barbeau (1915) and some unpublished work done by Lounsbury, the only systematic comparative work on the Iroquoian languages is Blin-Lagarde (1972). Lounsbury (1961) provides some glottochronological time depths for the historical branchings within this family, and points to some evidence which suggests an ancient split between Cherokee, Laurentian, Huron, and Tuscarora on the one hand and the Five Nations languages on the other — a split which was later superseded by the major branching into Northern and Southern Iroquoian.

1.3 1.3.0

THE STATE O F SIOUAN RESEARCH

Introduction

With respect, at least, to those languages which are still spoken or have been spoken in recent times, the languages of the Siouan family are somewhat more numerous and more widely distributed than those of the other two families discussed above. The most conspicuous Siouan language is Dakota (Sioux proper), which is still spoken in its various dialects by perhaps twenty thousand people of all ages on reservations in North and South Dakota, Nebraska, Montana, Minnesota, Manitoba, Saskatchewan, and Alberta. Iowa and Oto, two closely related dialects of a language which has sometimes been referred to as Chiwere, are still spoken by perhaps a few hundred older people in Oklahoma, Kansas, and Nebraska. A third Chiwere dialect, Missouri, is now extinct. A language closely related to Chiwere, and sometimes included under the same label, is Winnebago, of which there may still be a thousand or more speakers in Wisconsin and Nebraska. Those in Wisconsin may still include some children. Another language consisting of several dialects with distinct names is Dhegiha. Dialects still spoken in Oklahoma include Ponca, Osage, Kansa (Kaw), and Quapaw (Arkansas). There are no more than a few hundred speakers altogether. Kansa and Quapaw are the closest to extinction, if they are still spoken at all. Ponca, as well as a fifth Dhegiha language, Omaha, is also spoken in Nebraska, where Omaha remains the most viable language of this group with a thousand or more speakers, some of whom may still be children. Turning back to the north, we find the Crow language spoken by a few thousand people, including children, in Montana. The Hidatsa language in North Dakota has fewer than a thousand speakers, and perhaps no children

29 are learning it. On the same reservation in North Dakota there remains a diminishing handful of speakers of Mandan. Two Siouan languages, Ofo and Biloxi, were once spoken in Mississippi. Both languages became extinct in the early years of the present century. A number of Siouan languages were once spoken east of the Appalachians in Virginia and the Carolinas. The early historical situation is confused, and we have data from only three of these languages, all now extinct: Tutelo, Catawba, and Woccon. Tutelo and Catawba will be discussed below. Woccon is known to us only from a vocabulary of about 140 items first published in Lawson (1709: 225-30).

1.3.1

Dakota

Europeans first encountered the Dakota in the general area of the upper Mississippi River, in what is now Minnesota and western Wisconsin. Under pressure from other Indians, caused indirectly by white pressure further to the east, the Dakota gradually moved westward, displacing other tribes in turn. They spread across the Missouri River in the 18th century, and by the 19th century occupied a large territory in the states and provinces mentioned above, where their present-day reservations remain. There are usually said to be four major Dakota dialects: Santee (Dakota proper), Teton (Lakota), and Yankton and Assiniboine (collectively Nakota). Each has various minor subdialectal divisions. The Canadian variety of Assiniboine is called Stoney. The Dakota were first mentioned in recorded history by the Jesuits, in 1640, but the first mention which has any linguistic interest appears in the account by Louis Hennepin, a Recollet missionary who spent a few months among the Dakota in 1680. Hennepin wrote that he compiled a dictionary of the language, but it has not survived (see Cross 1938: 109). A century later Jonathan Carver, a Captain of provincial troops in the area during the 1760's, published an account of his observations, and included a vocabulary of Santee (Carver 1778: 433-40). Carver's material appears also in Barton (1797). Edward Umfreville, an 18th century fur trader, provided a short word list of Assiniboine (Umfreville 1790: [facing] 202), republished in Gallatin (1836: 374), as well as in Adelung-Vater (1806-17 3. 3: 263-5). The latter work also contains grammatical remarks on Dakota (256-64), under the early name Naudowessi (Nadowessier). Thomas Say of the Long expedition in 1819— 20 collected a Yankton vocabulary (Thwaites 1904-07 17: 290-8), also republished in Gallatin ( 1836: 307-67) along with a vocabulary of Dakota

30 proper compiled from several other sources. Gallatin (1836: 251-2) also included a few brief remarks on Dakota grammar from a General Cass. In the early 1830's Alexander Philipp Maximilian Prinz von Wied, collected vocabularies of Assiniboine, Yankton, and Teton, which are most easily found in Thwaites (1904-07 24: 215-7, 223-6). Horatio Hale collected a Dakota vocabulary and a shorter Yankton vocabulary which were published in Gallatin (1848: 83-9, 116). Catlin (1841 2: 262-5) also contains a Dakota vocabulary. The 19th century saw considerable Protestant missionary activity centered in the Minnesota area. One of the earliest of these missionaries was Jedediah Dwight Stevens, the author of a Sioux spelling book (Stevens 1836). Another, Samuel W. Pond, compiled a Hebrew-Dakota dictionary in 1842, with a view toward translating the Bible directly from Hebrew into Dakota. It was never published, but the manuscript is discussed by W. Gunther Plaut (1953). Certainly the most important of these missionaries from a linguistic point of view was Stephen Return Riggs, who worked among the Dakota for about 40 years beginning in 1837. He provided an abundance of material for the Indians in their own language, including a catechism, hymns, lesson books, the Minnesota Constitution, and The Pilgrim's Progress. In collaboration with Thomas S. Williamson he eventually translated the entire Bible into Dakota (Williamson - Riggs 1880). He also published a number of more specifically linguistic works. The first version of his Dakota grammar and dictionary appeared at mid-century (S. R. Riggs 1852), and an abridgment of the dictionary was prepared by his wife (M. A. C. Riggs 1852). The materials he collected were posthumously edited by J. Owen Dorsey in the form of an expanded Dakota-English dictionary (S. R. Riggs 1890) and a volume containing a grammar, texts, and ethnographic description (S. R. Riggs 1893). Several briefer works include a general discussion of the language (S. R. Riggs 1881a), as well as a short text (S. R. Riggs 1881b). The Riggs dictionaries, especially, remain of great value to linguists today despite certain shortcomings (above all his failure to distinguish between aspirated and unaspirated consonants). Other works from the 19th century include a grammar by Hans Georg Conon von der Gabelentz (1852a), vocabularies of both Teton and Assiniboine collected by Ferdinand Vandeveer Hayden (1863: 375-8, 389-91), a vocabulary prepared by two army lieutenants (Hyer - Starring 1866), and a general discussion of the language in Roehrig (1873), which suggested very tentatively that Dakota might be related to the Turanian (UralAltaic) languages. This speculation was improved upon by A. W. Williamson, the son of Thomas S. Williamson, who presented evidence

31 that Dakota was instead related to Indo-European (Williamson 1881, 1882). Lewis Henry Morgan (1871: 291-382) included kinship terms from nine Dakota dialects, the Santee obtained from Riggs and the rest collected by himself. Hale (1883c: 36-45) made use of Rigg's material. J. Owen Dorsey (1885) compared Dakota with several other Siouan languages. Other early sources on Dakota can be found in Pilling (1887). The Protestant missionary tradition begun in the 1830's produced its last important linguistic work with the publication of an English-Dakota dictionary by another son of Thomas S. Williamson (John P. Williamson 1902). Most of that work had been focused on the Santee dialect. Subsequently, however, the Teton dialect was given a great deal of attention by a Jesuit missionary, Eugene Buechel. He published a grammar of Teton (Buechel 1939), and left a vast amount of lexical material, which has just recently been published in dictionary form (Buechel 1970). It was, however, Franz Boas who was responsible for the most important linguistic work on Dakota in the present century. The first volume of the Handbook of American Indian languages contained a combined description of the Teton and Santee dialects. The Teton material was extracted by John R. Swanton from some manuscript texts, but the Santee was taken by Boas from Riggs's publications (BoasSwanton 1911). Most of Boas's later work was done in collaboration with Ella Deloria, a native speaker of Teton. The two authors produced a sketch of the language which appeared in I JAL in the early 1930's (Boas - Deloria 1933), and Deloria herself had published a major volume of texts in the preceding year (Deloria 1932). Boas subsequently wrote an article describing several features of the language which he found to be of general linguistic interest (Boas 1937). The culmination of this work, however, was the joint publication of a full-fledged Dakota grammar (Boas - Deloria 1941). This work undoubtedly stands as one of the high points in American Indian linguistics. Many years later Deloria saw to the publication of some brief textual material (Deloria 1954). Robert H. Lowie worked with the Assiniboine in the early years of this century. Lowie (1909: 263-70) has short texts with interlinear translation, and other texts, collected between 1907 and 1913, were published by his wife after his death (Lowie 1960c). There is a long Dakota word list in Fred M. Hans (1907: 309-58). James R. Walker (1914) contains Teton kinship terms, as well as a text in Teton dealing with the kinship system. Melvin R. Gilmore (1919: 139-47) has a list of Dakota plant names. G. Hubert Matthews (1955) gives a phonemic analysis of the speech of a single individual with an atypical linguistic

32 background. An attempt to delimit Dakota sentences and clauses by means of phonological and grammatical (non-semantic) criteria is described in Donald S. Stark (1962). Still more recently, a grammar of Assiniboine was published by Norman Balfour Levin (1964). Further work on Assiniboine (Stoney) has been done in Alberta by Warren Harbeck of the Summer Institute of Linguistics and by Allan R. Taylor. Paul WarCloud Grant (1971) is basically a Santee dictionary, with many Teton forms included. Taylor and David S. Rood have recently received funds for the preparation of pedagogically oriented materials on Teton.

1.3.2

Chiwere {Iowa, Oto, and Missouri)

The Iowa were a small group which, in earlier times, occupied various spots within the present state of Iowa and neighboring states. In 1836 they were given a reservation in Nebraska and Kansas. Some of them later settled in Oklahoma. The Oto were at first located south and west of the Iowa, near the confluence of the Platte River with the Missouri. For a time in the 19th century they occupied land in Nebraska and Kansas, but in the 1880's they moved to Oklahoma. The Missouri were once located on the Missouri River, near the point where it is joined by the Grand River in the state of Missouri. At the end of the 18th century they were badly defeated by the Sauk and Fox, and they suffered further in a war with the Osage in the early 19th century. From that time on, most of them lived with the Oto and moved with the latter to Oklahoma. It would appear that the earliest linguistic record of Chiwere is the brief Oto vocabulary collected by Thomas Say during the Long expedition of 1819-20 (Thwaites 1904-07 17: 290-8, 300). Say's material was also printed in Gallatin (1836: 307-67). Gallatin (1836: 377) also gives a very brief vocabulary of Iowa. A somewhat longer vocabulary of Oto was collected by Maximilian von Wied in the 1830's (Thwaites 190407 24: 285-93). Another brief vocabulary of Oto collected by Horatio Hale was printed in Gallatin (1848: 117). William Hamilton and Samuel M. Irvin were two Protestant missionaries who worked among the Chiwere tribes for many years during the 19th century. They published an Iowa spelling book and a 'primer' (Hamilton-Irvin 1843, 1849), but their most important linguistic work was an Iowa grammar (Hamilton-Irvin 1848). Hamilton also provided the remarks on the Iowa language contained in Schoolcraft (1851-57 4: 397-^06). Hayden (1863: 452-6) gives a vocabulary of "Iowa, or Oto". Morgan (1871: 291-

33 382) lists two sets of kinship terms — for Iowa and for Oto-Missouri. J. Owen Dorsey did considerable work with Chiwere in the latter part of the 19th century, but most of it remains unpublished in the Smithsonian archives. Reference can be made to a brief published text in Oto with interlinear translation (Dorsey 1880-81), and to the Chiwere forms in Dorsey (1885). Other 19th century sources are listed in Pilling (1887). The only more recent source is the very brief grammar of IowaOto by William Whitman ( 1947). Whitman's material is used for examples in Voegelin (1947).

1,3.3

Winnebago

Closely related linguistically to Chiwere, the Winnebago (often called the Hochangara in the early literature) once lived south of Green Bay in what is now Wisconsin. Some have remained in that state while others, after many changes of location, eventually settled on a reservation in northeastern Nebraska. There is not much early material on Winnebago. A brief vocabulary taken down by Major Stephen H. Long in 1817 can be found in Thwaites (1904-07 17: 306-8). Nicholas Boilvin, an Indian agent, collected a Winnebago vocabulary sometime before his death in 1824. It, together with material from other sources, is given in Gallatin (1836: 307-67). Another brief vocabulary obtained by Horatio Hale appears in Gallatin (1848: 116). Ferdinand Vandeveer Hayden (1869: 411-21) collected and published some "grammatical forms and phrases" as well as a vocabulary. Morgan (1871: 291-382) gives Winnebago kinship terms. Most of J. Owen Dorsey's material remains in manuscript, but some of it can be found in Dorsey (1885). A few other early sources are mentioned in Pilling (1887). Melvin R. Gilmore (1919: 139-45, 148-9) lists a number of Winnebago names for plants. The ethnologist Paul Radin worked among the Winnebago from 1908 to 1913. His linguistic material was used by Boas in presenting comparative remarks on Winnebago in Boas-Swanton (1911) on Dakota. Radin (1923) contains some ritual speeches in Winnebago with free translations, and there are many Winnebago words scattered throughout the work. Much later, two monographs were published containing texts in Winnebago accompanied by extensive commentaries (Radin 1949, 1950). Radin's notes also formed the basis for the Winnebago dictionary recently prepared as a dissertation by Mary Marino (1968). Several other works appeared during the 1940's.

34 One was a brief article on word play in Winnebago by Amelia Susman (1941), incidental to a larger but unpublished dissertation (Susman 1943). Some of Susman's data were utilized in Hockett (1942). Three years later there appeared a small but useful grammar written by William Lipkind (1945). Two short Winnebago texts were published by Thomas A. Sebeok shortly after (Sebeok 1947). Besides the Susman and Marino dissertations mentioned above, a recent dissertation on Winnebago phonology was completed by Anita Marten (1964).

1.3.4

Dhegiha (Omaha, Ponca, Kansa, Osage, Quapaw)

In prehistoric times the ancestors of the Dhegiha speakers may have lived farther to the east (according to tradition, near the junction of the Wabash and Ohio Rivers), but earliest European contact found them in the central plains. The Omaha and Ponca were on the southwest bank of the Missouri River in northeastern Nebraska, the Kansa on the Kansas River in the present state of Kansas, the Osage on the river of that name in Missouri, and the Quapaw near the junction of the Arkansas River with the Mississippi. The Omaha still live in Nebraska, as do some of the Ponca, but the majority of the latter have been in Oklahoma since 1877. The Kansa have been in Oklahoma since 1873. During much of the 19th century the Osage were located chiefly in Kansas, but in the early 1870's they were established on a reservation in Oklahoma where they now live. The Quapaw occupied various spots in Arkansas, Kansas, and Oklahoma, until they were finally restricted to a small area in northeastern Oklahoma in 1867. Perhaps the earliest published record of a Dhegiha dialect is the vocabulary of Quapaw (Arkansas) in the second edition of Benjamin Smith Barton (1798), collected by someone named Bossu. From the beginning of the 19th century there are several vocabularies of Osage: one in Adelung - Vater (1806-17 3. 3: 273-4) collected by Albert Pike, another in Bradbury (1817: 213-9), and still another, collected by one Dr. Murray, in Vater (1821: 53-62). Both the Bradbury and the Murray materials were used in Gallatin (1836: 307-67). In the 1830's Maximilian von Wied collected vocabularies of Kansa, Omaha, Ponca, and Osage (Thwaites 1904-07 24 : 229, 280-5, 294, 296-300). At about the same time two missionaries, William B. Montgomery and W. C. Requa, published The Osage first book (Montgomery-Requa 1834). Gallatin (1848: 83-9) contains some Osage words collected by Horatio Hale. In 1819-20 Thomas Say of the Long expedition took down vocabularies

35 of Kansa and Omaha (Thwaites 1904-07 17: 290-8, 301-3), which also appear in Gallatin (1836: 307-67). Hale's very brief vocabularies of Omaha and Quapaw appear in Gallatin (1848: 117). Edward McKenney, another missionary, prepared an Omaha primer (McKenney 1850). Hayden (1863: 448-52) also has an Omaha vocabulary. Morgan (1871 : 291-382) includes kinship terms for Ponca, Omaha, Kaw (Kansa), and Osage-Quapaw. Other early sources are mentioned in Pilling (1887). A great deal of work was done on the Dhegiha dialects in the latter part of the 19th century by J. Owen Dorsey, who was in fact responsible for grouping them together under that name. His most voluminous published work (Dorsey 1890) consists of texts in Omaha and Ponca, with interlinear and free translations, and notes. Much earlier, while still a missionary, he prepared a Ponca primer (Dorsey 1873). Omaha texts with interlinear and free translations can be found in Dorsey (1879-80) and Dorsey (1881), and Osage texts are given similar treatment in Dorsey (1888). Dorsey (1884) contains a number of Omaha words and sentences. Dorsey (1885) includes comparative phonological information and vocabulary from Ponca, Kansa, and Osage. References to other publications and manuscripts of Dorsey can be found in Pilling (1887: 24-6) and Dorsey (1890: xvii-xviii). Working from Dorsey's materials, particularly Dorsey (1890), Franz Boas wrote an article on Ponca grammar (Boas 1907), and interspersed remarks on Ponca (as well as Winnebago) in Boas-Swanton (1911). The large ethnographic description of the Omaha by Alice C. Fletcher and Francis La Flesche contains a number of Omaha words scattered throughout (Fletcher-La Flesche 1911). There is an ethnobotanical study of Omaha by Melvin R. Gilmore (1913), plus a list of Omaha plant names in Gilmore (1919: 139-45, 147-8). More recently there has been an article by Nils M. Holmer (1945) on certain aspects of the Ponca-Omaha and related sound systems. Howard (1965) provides a number of Ponca terms. Osage has been documented in this century, first through the dictionary by La Flesche (1932), and later by two articles on Osage phonology and morphology by Hans Wolff (1952). Wolff also discussed attempts by his informant to write Osage words (Wolff 1958).

1.3.5

Crow

The Crow (earlier often called Upsaroka) have, in historic times, always been located near the Yellowstone River in Montana, their present reservation being on the Big Horn River, a southern tributary of the Yellowstone.

36 Perhaps the earliest recording of the Crow language that can be found in a published source is the very brief vocabulary collected by Thomas Say of the Long expedition in 1819-20 (Thwaites 1904-07 17: 299). It was reprinted in Gallatin (1836: 377). Another brief vocabulary was collected by Maximilian von Wied (Thwaites 1904-07 24: 222). Gallatin (1848: 83-9) gives a somewhat longer list obtained by Horatio Hale. Ferdinand Vandeveer Hayden (1863: 395-420) presents not only a vocabulary, but also some remarks on the grammar. Morgan (1871: 291-382) includes Crow kinship terms. For other early sources see Pilling (1887: 22). Recent work on Crow is somewhat more extensive than that on most of the other Siouan languages. Much of it was done by the ethnologist Robert H. Lowie, who paid a great deal of attention to the language. His most important linguistic works include a text accompanied by grammatical notes (Lowie 1930), a grammatical sketch followed by a text and analysis thereof (Lowie 1941), and a large volume of posthumously published texts accompanied by Crow-English and EnglishCrow vocabularies in a separate volume (Lowie 1960a, 1960b). Lowie had a special interest in literary style in Crow, in which connection reference can be made to Lowie (1932, 1950, 1959). Still more recent work on Crow has been done by Dorothea V. Kaschube. Her principal work is a grammar (Kaschube 1967). An earlier article on tones in Crow (Kaschube 1954) was followed by a restatement by Eric P. Hamp (1958), and still another interpretation, in terms of ordered rules, by G. Hubert Matthews (1959a). Ray Gordon of the Summer Institute of Linguistics has been working on Crow for a number of years.

1.3.6

Hidatsa

The Hidatsa, formerly often called the Minitari and locally still referred to as the Gros Venires, have always, within historic times, been located within what is now the state of North Dakota, at various points along the Missouri River. Having been earlier slightly farther south, in 1845 they moved to the Fort Berthold area, where they still are. The earliest published vocabulary of Hidatsa again appears to be that collected in 1819-20 by Thomas Say, now available in Thwaites (190407 17: 290-8, 304). Maximilian von Wied contains a Hidatsa vocabulary collected in the 1830's (Thwaites 1904-07 24: 261-76). Gallatin (1836: 307-67) repeats the material from Say, and a short word list obtained by

37 Horatio Hale is published in Gallatin (1848: 117). Schoolcraft (185157 3: 256) also repeats some of Say's material. Another vocabulary of Hidatsa appears in Hayden (1863: 424-6). Morgan (1871: 291-382) includes Hidatsa kinship terms. Other early sources are listed in Pilling (1887: 35, 51). Of considerably more importance is the work of Washington Matthews, an army surgeon who was stationed in the Hidatsa area between 1865 and 1872. Matthews gave a great deal of time and attention to the Hidatsa language, and wrote up what he found. His principal publication is the grammar and dictionary contained in Matthews (1877), compiled and revised from his earlier grammar and Hidatsa-English dictionary (Matthews 1873) and English-Hidatsa dictionary (Matthews 1874). His material was used in Horatio Hale (1883c: 36-45). Robert H. Lowie collected a number of Hidatsa texts in 1911. Almost thirty years later this material was linguistically annotated and published by Zellig Harris and C. F. Voegelin (Lowie 1939). R. H. Stetson (1946) subsequently wrote about certain aspects of Hidatsa phonetics. Work done at the 1953 Linguistic Institute led to a discussion by C. F. Voegelin and Florence M. Robinett (1954) of what they called Hidatsa "mother language" — the manner in which a speaker may modify his pronunciation "for the benefit of children or of strangers struggling to acquire the language" (Voegelin-Robinett 1954: 65). From a present-day point of view it is interesting that they found such speech to "coincide with or resemble" (Voegelin-Robinett 1954: 70) their tentative morphophonemic representations. Shortly thereafter Robinett (1955) published a three installment description of the language. More recently there has appeared an extensive transformational treatment of Hidatsa written by G. Hubert Matthews (1965). The first chapter of that work includes a useful critical discussion of the earlier works by Washington Matthews, HarrisVoegelin, and Robinett.

1.3.7

Mandan

In historic times the Mandan have lived in approximately the same area of North Dakota as the Hidatsa, and are presently located with the latter (as well as the Arikara) on the Fort Berthold Reservation. They were first encountered by Verendrye in 1783, and were described in some detail by Maximilian von Wied and Catlin in the 19th century. They attracted special attention because they were said to have a lighter skin color and to differ in other ways from other Indians, and they were thus

38 supposed by some to have had a European origin. The first published Mandan vocabulary may be that in C. S. Rafinesque (1832-33), obtained principally from George Shannon, said to have been a member of the Lewis and Clark expedition. In the 1830's Maximilian von Wied gave more attention to Mandan than to any of the other languages he recorded, and he subsequently published not only a vocabulary but also a small grammar (Thwaites 1904-07 24 : 234-61). Catlin (1841: 261-5) includes a Mandan vocabulary, and compares a few Mandan and Welsh words (with the idea that the Mandan were descended from displaced Welshmen). A Mandan vocabulary provided by James Kipp was published in Schoolcraft (1851-57 3: 255-6, 446-59). Hayden ( 1863:435^4) also includes grammatical remarks as well as a vocabulary. Mandan kinship terms appear in Morgan (1871: 291-382). Other early sources are given in Pilling (1887: 48). Early in the present century two Harvard students put together a description of Mandan, including a grammatical sketch and a vocabulary, based on the various published sources then available (Will - Spinden 1906: 188-219). Edward Kennard later published a grammar of Mandan (Kennard 1936) which until recently has been the principal modern source of information on the language. However, a Mandan dictionary accompanied by considerable phonological and grammatical information has now been completed (Hollow 1970).

1.3.8

Ofo and Biloxi

The Ofo were located historically on the Yazoo River in the present state of Mississippi. After various sojourns on or near the Mississippi River, they seem to have been entirely lost from history between the years 1784 and 1908, when a single Ofo survivor was found by John R. Swanton living among the remnants of the Tunica in Louisiana (Swanton 1909). The Biloxi were once located on the lower Pascagoula River in Mississippi. They subsequently lived in several locations in Louisiana, and some of them in Texas and Oklahoma. They too seem to have dropped out of sight for a time, until in 1886 a group of them was discovered by Albert S. Gatschet in Louisiana. Gatschet collected some linguistic material, evidently the first to have been recorded for either of these languages. Gatschet's material itself has never been published, but it was enough to enable the linguists at the Bureau of American Ethnology to determine the Siouan affiliation of Biloxi, and it stimulated J. Owen Dorsey to

39 visit these people for extensive linguistic work in 1892 and 1893. His report on the Biloxi included a grammatical sketch of the language (Dorsey 1894). The material which he left at his death two years later was subsequently edited by Swanton, and published in a single volume together with Swanton's own material on Ofo (Dorsey-Swanton 1912). In addition to Biloxi and Ofo dictionaries, this work contains a number of annotated texts in Biloxi, and a collection of Biloxi sentences. It formed the basis for several later publications. Voegelin (1939) discusses sound correspondences between Ofo and Biloxi. Holmer (1947) talks about the development of a prothetic vowel in Ofo (as well as Biloxi and Tutelo). Haas (1969) provides several kinds of advice for users of the Dorsey - Swanton volume. Swanton's Ofo material was all that was ever recorded for that langugage. For Biloxi, however, Mary Haas and Morris Swadesh collected some 54 words in 1934 from a woman living in Texas who had evidently been an able speaker of the language many years earlier. This material has recently been published (Haas 1968). Paula F. Einaudi is preparing a grammar of Biloxi based on Dorsey's and Gatschet's published and unpublished materials.

1.3.9

Tutelo

In the 17th century the Tutelo were living in western Virginia. They subsequently moved northward, and in 1753 were adopted by the Cayuga in New York State, with whom they moved to Canada after the Revolution. The principal source of information on the language is the article by Horatio Hale (1883c), based on work which he did with the last fluent speaker of the language in 1870, supplemented by material subsequently obtained from others who remembered the language to some degree. In the early years of the present century a few scattered words were resurrected from the memory of people who had once heard the language (Sapir 1913, Frachtenberg 1913). A few kinship terms recorded in the early 1930's by Frank G. Speck are given in Speck - Schaeffer (1942: 573-4).

1.3.10

Catawba

Since European contact, at least, the Catawba have always been located in an area centered in northern South Carolina. They were already known to Europeans in the 16th century, and during the early colonization

40 period they and the Cherokee were the two most important Indian groups of the Carolinas. During the 18th century, however, wars and smallpox reduced them to a state of comparative insignificance. They have for many years occupied a tiny reservation near Rock Hill, South Carolina, but the language is no longer spoken. Some Catawba words appear in Barton (1798), the second edition of Barton (1797). This material as well as a larger vocabulary collected by one J. L. Miller was included in Gallatin (1836: 307-67). Another and still longer word list is to be found in Lieber (1858), along with a few grammatical notes. Albert S. Gatschet worked with the language in 1881, and published a grammatical sketch some years later (Gatschet 1900). Gatschet (1902) also contains a few Catawba words. John R. Swanton did linguistic work on Catawba in 1918, and Swanton (1918) contains some ethnobotanical information. Frank G. Speck worked with Catawba informants intermittently between 1913 and 1944. Texts collected in 1913 from a Catawba woman living among the Cherokee were published with interlinear and free translations in Speck (1913). Speck (1924) gives a few Catawba words tending to confirm the notion that there were Siouan peoples on the coast of South Carolina in the 16th century. More texts were published in Speck (1934), and a last brief one, obtained from the man who was probably the last fluent informant, appeared in Speck (1946). Speck - Schaeffer (1942) is based on kinship terms from several sources, including Speck's own material. Frank T. Siebert, Jr., worked with the last few Catawba speakers in 1941. His two-part comparative article (Siebert 1945) cites some of the material which he collected, more of which remains in manuscript. More recently, G. Hubert Matthews has worked with a man who once spoke some Catawba with his grandfather and a cousin. Matthews-Red Thunder Cloud (1967) discusses the phonology and morphology of material obtained from this informant, and gives some texts and a word list.

1.3.11

Relationships within Siouan

The relatedness of the western Siouan languages was the first to be recognized. With the exception of Winnebago they occupied a geographically contiguous area, and their similarities were readily apparent.. In Gallatin's classification (Gallatin 1836: 120-8, 306) they were grouped together under the name Sioux. (He also hesitantly included Cheyenne, whose Algonquian affiliation was not then recognized.) Nothing was known of the Biloxi and Ofo languages at that time. When

41 Biloxi was first recorded by Gatschet in 1886 its Siouan nature was immediately seen, and the same can be said of Ofo when it was recorded by Swanton in 1908. As for the eastern languages, Hale's description of Tutelo emphasized the Siouan affiliation of that language, and he included a comparative vocabulary of Tutelo, Dakota, and Hidatsa (Hale 1883c: 36-45). The Siouan nature of Catawba was not so obvious. A good discussion of the history of this problem is available in Siebert (1945: 100-1). Gallatin (1836: 87, 306) listed Catawba as an independent language family, including also Woccon as originally suggested by AdelungVater (1806-17 3. 3: 308). Although the staff of the Bureau of American Ethnology credited Gatschet with being the first to discover that Catawba was related to Siouan (e.g. Powell 1891: 112), Siebert points out that the credit should really go to Lewis Henry Morgan, although Morgan did not support his suggestion (Morgan 1870: 54) with concrete evidence. Perhaps, however, Latham (1860: 327) should be given some credit also. A brief work by A. F. Chamberlain (1888) made use of the same Gallatin material on which Morgan had based his conclusion. But no extensive verification of this hypothesis appeared until Siebert (1945) settled the matter. A useful survey of the work that has been done to establish the existence of Siouan languages in the East is Sturtevant (1958). The early subgroupings of the Siouan languages were based as much on geographical as on linguistic criteria. Gallatin (1836: 120) mentioned four subdivisions: Winnebago, Sioux proper, the "Minetare group" (Crow, Hidatsa, and Mandan), and Osage "and other southern kindred tribes". With more languages recognized, and with considerably more data available, Swanton (1923) suggested a classification of Siouan into four subgroups: (1) a northeastern group including Hidatsa, Dakota, Biloxi, Ofo, and Tutelo; (2) a southeastern group including Catawba and Woccon ; (3) a southwestern group including the Dhegiha dialects ; and (4) a northwestern group including the Chiwere dialects and Winnebago. He purposely left Mandan outside this subgrouping, being uncertain of its proper assignment. Swanton (1936) emphasized the fact that Tutelo and Catawba belonged to different subgroups. Voegelin (1939) strengthened the idea that Biloxi and Ofo belong within the same subgroup, and Voegelin (1941a) did the same for Crow and Hidatsa. Pierce (1954) also discussed the degree of relatedness of Crow and Hidatsa. The overall subgrouping proposed in Voegelin (1941b) has become the standard generally referred to. Voegelin also established four subgroups, but they differed in certain respects from Swanton's. They included: (1) an eastern group consisting only of Catawba; (2) a so-called

42 Ohio valley group consisting of Ofo, Biloxi, and Tutelo (with the idea that the prehistoric location of these peoples was in the Ohio valley area) ; (3) a Missouri River group consisting of Hidatsa and Crow ; and (4) a Mississippi River group consisting of Chiwere, Winnebago, Dhegiha, Dakota, and Mandan. The presence of Mandan in the last group seems questionable, however ; at the present time it would seem preferable to place Mandan in a subgroup by itself, although one that is more closely tied to the other Siouan languages than is Catawba. In contrast to the Caddoan and Iroquoian families, where the amount of comparative work to date has been minimal, several scholars have paid a great deal of attention to comparative Siouan. The earliest was J. Owen Dorsey, who published a comparative phonology of Dakota, Dhegiha, Chiwere, and Winnebago (Dorsey 1885). A more recent and more thorough study is Wolff (1950-51). G. Hubert Matthews has also worked in this area, and has written a dissertation on the subject (Matthews 1958) as well as an article dealing with the reconstruction of the Proto-Siouan kinship terminology (Matthews 1959b). Terrence S. Kaufman has also done unpublished work on comparative Siouan. It can hardly be said, however, that all the problems in this area have been laid to rest. The work accomplished so far has clearly been only a prelude to what might eventually be done within this diverse and interesting language family.

REMOTE RELATIONSHIPS

2.1 2.1.0

THE MACRO-SIOUAN HYPOTHESIS

Introduction

From time to time there have been suggestions that the three language families discussed in the preceding three sections are remotely related to each other. At the beginning of this volume the name 'Macro-Siouan' was suggested for this hypothetical linguistic stock. In what follows, the possible relationships of Siouan and Iroquoian, of Siouan and Caddoan, and of Iroquoian and Caddoan will be discussed in that order.

2.1.1

Siouan and Iroquo ian

It is this relationship which is the best documented of the three, as might be expected from the fact that the Siouan and Iroquoian languages themselves have been better documented than the Caddoan. The possibility of such a relationship was perhaps implied in Latham (1846 : 44), but it was explicitly stated in Latham (1860: 327). Morgan (1871: 150-1) asserted his belief that the Iroquois were "an early offshoot" of the Sioux. Edward Sapir, in his famous 1929 proposal concerning the classification of North American languages, placed both Iroquoian and Siouan in his large Hokan-Siouan grouping, but he did not posit any special closeness of relationship between them (Sapir 1951: 173). Shortly thereafter, however, Louis Allen published the first real evidence supportingthe Siouan-Iroquoian relationship (Allen 1931). More recently the hypothesis has been strengthened by Chafe (1964a), who listed 67 possible cognates with tentative Proto-Siouan-Iroquoian reconstructions and a tabulation of phonological correspondences.

44 2.1.2

Siouan and Caddoan

Latham ( 1860: 327) tentatively suggested a Siouan-Caddoan connection, and again Sapir (1951: 173) placed both families within Hokan-Siouan without positing any special closeness of relationship. I have assembled the following evidence in support of a Siouan-Caddoan relationship. It is not possible at present to set down a list of hypothetical cognates which would be subject to explanation on the basis of a plausible Proto-Siouan-Caddoan phonological system. Some lexical resemblances have indeed been noted, but they are not of the quantity or nature that would permit one to forcefully reject the explanation that they are fortuitous (cf. English cut and Caddo kat 'knife')· Then too, there are some resemblances between Caddo and Siouan which are in all likelihood the result of diffusion, probably from Osage (La Flesche 1932) into Caddo:

'mother' 'deer'

Caddo Hna ? da?

Osage ina ta

Resemblances such as the following are tantalizing, if inconclusive :

'bird' 'blood' 'arrow' 'earth' 'man'

Caddo Winnebago banit wanik bah ?uh wa9ih ba? m4 wádat mq wit 'self (Pawnee pita 'man')

Dakota we wa-(h(kpe) mq-(kha) wichá

More convincing are facts like the following. A well-known feature of Siouan languages is their use of "instrumental" prefixes in the derivation of verb bases. An instrumental prefix is typically represented by a consonant-vowel syllable ; for example, Winnebago nq-, mq-, wa-, gi-, ra-, ru-, ta-, bo-. It usually carries a meaning which is in some sense instrumental. The meaning may be generalized and impersonal, such as 'by heat' or 'by cutting', or it may be specific and personal, such as 'with the mouth' or 'with the foot' ; there are in fact two distributionally distinct classes which roughly reflect this semantic distinction. As would be expected, many of the meanings associated with particular combinations of these prefixes with verb roots are idiomatic : not predictable from any consistent meaning assignable to either the root

45 or the prefix. The following examples from Winnebago (Lipkind 1945) show some of the occurrences of the prefix gi-, said to have the meaning 'by striking' : gi-sak gi-kúnuk gi-pére gi-hiri gi-zé gi-xúx gi-xóro gi-gás gi-sára

'to 'to 'to 'to 'to 'to 'to 'to 'to

In Caddo most verb bases contain two formally distinguishable parts, which I will refer to as the preverb and the verb root. Examples are ka-dis 'to wash', ya-9ah 'to be', yi-bahw 'to see'. Usually these two parts are contiguous, but there is a 'plural' element (usually -wa-; see below) which intrudes between them : compare hákkadíssa 9 'he is washing' with hákkawadíssa7 'they are washing'. With a few exceptions, one particular verb root will occur always and only with one particular preverb. A preverb is typically represented by a consonant-vowel syllable; for example: bi-, ka-, ki-, na-, ni-, pa-, ya-, yi-, 7a-. There are about as many of them as there are instrumental prefixes in Siouan languages. For the most part these Caddo preverbs must be accepted as simply arbitrary appendages to the verb roots. It is usually problematic whether they can be associated with any consistent semantic feature. The preverb yi-, for example, is found in such bases as the following : yi-bahw yi-Suki yi-yah yi- ?a ?n

'to 'to 'to 'to

see' write' proceed' plant'

There is at least one case, however, in which there does seem to be an identifiable common element of meaning among many of the bases which contain a specific preverb : ki-c'ákPni ki-c'ud ki-duk ki-k'as

'to 'to 'to 'to

chew' peel' bite, break' shell' (e.g. corn)

46 ki-k'úd ki-náh 9y ki-paáhnu ki-saki ki-sáw? ki-siid ki-wawik

'to suck' 'to cut' 'to scratch oneself 'to pound, mash, crack' 'to scrape' 'to sew' 'to tear'

9

All these bases have to do with inflicting some sort of violence with an instrument, either part of one's body or a separate tool. The coincidence with the Winnebago examples cited earlier is threefold : there is a clear semantic resemblance (it would be hard to say how the common semantic element in these Caddo examples differs from that in the Winnebago examples) ; there is a correspondence in position (direct prefixation to the verb root); and there is a similarity in shape. Winnebago gi- is, in fact, apparently a reflex of Proto-Siouan *ki-. (The Winnebago vowel is anomalous ; other Siouan prefixes said to have the meaning 'by striking' go back to a Proto-Siouan *ka-.) At least one more parallel of the same kind can be mentioned. There is a Siouan instrumental prefix reconstructable as *ra- which has the general meaning 'with fire or heat'. It appears in Winnebago as ta-; for example (Lipkind 1945 : 20): ta-xú ta-xére ta-jók ta-wús

'to 'to 'to 'to

burn' fry' cook till tender' dry'

Caddo has several bases containing the preverb na- (alternating with ta- in initial position) in which the meaning of 'fire' or 'heat' is also involved : na-bahn na-hak na-hasi

'to catch fire' 'to dry' 'to be cooked'

Although there are fewer Caddo examples available, the coincidence here seems to parallel that which exists in the case of ki-. There is, however, a complication in the Caddo forms. Internal evidence shows that the preverb in these bases developed out of an incorporated noun root nak- 'fire'. There are other bases in which this noun root is the first

47 element; for example, nak-hàh 'to burn'. In the above examples one can identify a preverb na- only on the basis of the plural forms, since it is only in the plural that the difference between na- and nak is clear. The first element in these examples has evidently been reshaped from nakto na- in the plural, probably on the model of numerous other bases that begin with one of two other elements having the shape na-. This derivation of Caddo na- from nak- does not necessarily vitiate the parallel to Siouan, however. It may, on the contrary, be a useful clue to the way these instrumental prefixes came into being. It is certainly possible, although probably demonstrable only through such indirect shreds of evidence as this, that the instrumental prefixes were once incorporated noun roots which took on specialized function and perhaps underwent reshaping in the process. Another parallel between Caddoan and Siouan languages is the common occurrence within verbs of an element that specifies whether a lying, sitting, or standing posture is involved in the verb's total meaning. This element is prefixed to the verb root in Caddoan, but suffixed in Siouan. The following are possible reconstructions of the Siouan forms. There is probably a morpheme boundary between the syllables (some languages show reflexes of the first syllable only) : *-wâki *-ráki *-hàki

'lying' 'sitting' 'standing'

The Caddo element which has the most suggestively similar shape is ? aniki- 'standing'. In Caddo, too, it would appear that the -ki- was at one time a separate element ; cf. the other two forms, ?ini- 'lying' and ?awi'sitting'. Thus there is agreement of the two language families in specifying these three postures within verbs, in the shape of the Siouan forms when compared with Caddo 'standing', and in the ancient segmentability of these comparable forms.

2.1.3

Iroquoian and Caddoan

Latham (1846: 44) suggested more specifically that Caddo is related to the Iroquoian languages, and Sapir (1951: 173) placed Iroquoian and Caddoan together in a special subgroup within Hokan-Siouan. He never made clear his reasons for doing so, however, and this relationship, like that between Siouan and Caddoan, has remained undocumented up to

48 the present time. The following evidence for an Iroquoian-Caddoan relationship is taken largely from the Caddo and Seneca languages, but the implication is that the features described for these two are retentions from Proto-Caddoan and Proto-Iroquoian respectively. There are just a few lexical resemblances which can be cited : 'to pound corn' (verb root) 'to make' (verb root) 'to dye' (verb root) 'faeces' (noun or noun root)

Caddo (na)-dá 7 ( ?a)- '?nih (naca)-sú? 9 idah

Seneca -the ?t-çni-(ah)-so-i ?ta-

Perhaps more will appear as work on the Caddoan languages progresses, but for the moment the following kinds of evidence are more substantial. A somewhat oversimplified but useful generalization about both Caddo and Seneca verbs is that they consist of four major parts. The first part, counting positionally from left to right, is a motley collection of prefixes and prefix combinations involving meanings of tense, aspect, subordination, location, relation, negation, and so on. The second part is a group of personal prefixes which relate to the subject or object of the verb, or to a combination of subject and object. The third part is a verb base, which may include anything from a simple verb root to an elaborate combination of compounded roots and modifiers. The fourth part is a relatively small set of suffixes expressing aspect or tense or, in Caddo, location. Let us concentrate here on the personal prefixes and on certain items immediately adjacent to them in the verb base. We can begin by comparing those prefixes which are associated with singular subjects and objects in the various persons. For Caddo we will consider only the 'real' prefixes; there is a parallel 'unreal' set which is used in negations, questions, conditions, and elsewhere. Caddo does not overtly express third person, but its 'indefinite' person relates suggestively to one of the Seneca third persons :

'first person' 'second person' 'indefinite' (Caddo)/ 'feminine-indefinite' (Seneca) 'neuter' 'masculine'

Caddo Subject Object kuciyah?si-

Seneca Subject Object wak(e)k(e)sas(e)-

yi-

ye-

yu-

kaha-

(ya)koyoho-

49 (Seneca (e) is subject to different morphophonemic treatment than e, and (ya) is lost in word-initial position. These facts are not relevant to the hypothesis presented here, and must be explained on some other grounds.) We might imagine a regular model such as the following as a basis for stating the Caddo-Seneca similarities :

'first person' 'second person' 'indefinite'

Subject keseye-

Object kosoyo-

Not that an arrangement of precisely this form need be assumed to have existed in the remote past. For the purpose mentioned it is a convenient arrangement to work from, but it may overlook complications which are here extraneous. For Caddo we need state only the following developments (equating e of the model with Caddo i, and o with u ; Caddo has only the vowels i, u, and a) : (1) k > c before the front vowel. (2) By a "push-chain" development an innovational 'second person subject' form yahdisplaced si-, which in turn displaced su-, which in turn was lost. For Seneca the following statements are necessary : (1) Two new ways of representing the 'objective case' arose; specifically, the prefixing of wa- in the 'first person' and the use of a rather than o in the 'second person'. In the 'first person object' form, however, other Iroquoian languages (including at least Cayuga, Oneida, Tuscarora, and Cherokee) show an anomalous w after the k in some environments, suggesting that an earlier shape of the Iroquoian 'first person object' was *kw- (Lounsbury 1953: 69). (2) The meaning 'indefinite' was extended to include 'feminine', while 'neuter' and 'masculine' prefixes were added to form the following intermediate arrangement :

'feminine-indefinite' 'neuter' 'masculine'

Subject yekaha-

Object yokoho-

50 It might alternatively be posited that these three third person genders were originally present and were lost in Caddo(an). The reasons they are treated here as innovations in Seneca (or Northern Iroquoian) are :

(a) The retention of the 'indefinite' meaning alongside the 'femine' in Iroquoian, suggesting that the three genders were not always on a par. (b) The different vowel in the innovational subject forms. (c) The absence of the 'neuter' category outside the singular, where it fuses with 'feminine-indefinite'. (d) The absence of all third person gender distinctions in Cherokee.

(3) An interchange of 'feminine-indefinite' and 'neuter' object forms, but with retention of (ya) < yo as part of the 'feminine-indefinite' object. The plausibility and simplicity of this model intriguingly suggest a common origin for these Caddo and Seneca prefixes. In both Caddo and Seneca it is possible to specify 'dual' or 'plural' number for the subject or object of the verb. (The same device is used for the number of both subject and object, and sometimes there can be ambiguity as to which personal referent a number morpheme affects.) 'Plurality' is expressed in Caddo by a syllable -wa- that usually occurs between the preverb and the verb root. Thus, yi-bahw means 'to see' and yi-wa-bahw is the same base with specification of a plural subject or object. In Seneca 'plurality' is also expressed by a syllable -wa-. Seneca verb bases do not have the two parts described above for Caddo verbs, but consist of either a single root or a root with various prefixed and/or suffixed extensions. The syllable -wa- occurs between the personal prefix and the total verb base, whatever the latter may consist of ; thus, se-kçh 'you see' and s-wa-kçh 'you (plural) see' or 'it sees you (plural)'. Here again, with this plural morpheme, there is a clear similarity between meaning, form, and function in Caddo and Seneca. The Caddo 'dual' indicator does not occur in the same position as the 'plural', but in a position analogous to both the 'dual' and 'plural' in Seneca; that is, directly following the personal prefix. The Caddo 'dual', however, is -wiht-, while the Seneca 'dual' is -ni-. Just possibly Caddo -wiht-, which alternates with piht- initially, is relatable to the Siouan 'plural' suffix *-pi. It may be noted that Caddoan postural prefixes ('lying', 'sitting', 'standing') also correspond to Siouan suffixes.

51

In addition to the prefixes mentioned above, Caddo shows in some verbs a prefix from the following set : hanihakakani-

'nonsingular subject' 'nonsingular object' 'indefinite transitive' (involvement of the indefinite category in a subjectobject combination)

Such a prefix occurs between the personal prefix and the verb base. Each has some peculiarities associated with its occurrence, but all that needs to be said here is that these three forms are evidently relics of an earlier prefix system — perhaps one that occurred only with verbs of a certain kind — on which the modern system has been superimposed. A relationship of these prefixes in Caddo to the following Seneca personal prefixes is suggested : hçn- (alternating with ha ti-) hakç-

khni- < *kani-

'masculine plural subject' 'nonsingular object acted upon by masculine singular subject' 'nonmasculine dual subject'

It might be further speculated that the ha- in the Caddo forms bears an ultimate relationship to Seneca ha- 'masculine subject', Caddo ka- to Seneca ka- 'neuter or nonmasculine subject', Caddo -ni- to Seneca -ni'dual', and Caddo -ka- to an Iroquoian indicator of nonsingularity — an element ka- which is prefixed to personal forms in Cayuga, Tuscarora, and Cherokee (Lounsbury 1961: 12-3). It most often affects the number of the object of a transitive prefix; note that Caddo haka- indicates 'nonsingular object'. There is a syllable 9i- which occurs before some Caddo noun and verb roots, and whose only function seems to be to protect the left flank of the root from occurring initially in a word or from coming into direct contact with a personal prefix. For example, the verb root wayuh 'to climb' (with a variant shape -wyuh) will tolerate an immediately preceding plural, as in pawyuhah 'they climb', but adds 7i- to protect itself initially ( Hwyuhah 'he climbs') and from a pronominal prefix (ci-wayuhah < *ci?iwayuhah Ί climb'). There is also in Seneca a syllable ?i- which occurs at the beginning of some verbs, and whose only function is to prevent the occurrence of a verb that would contain only one vowel. Thus,

52 one would otherwise expect the imperative 'hit it!' to be *jçt, but it is in fact ?ijçt (Chafe 1967: 14). This use of ?i- as a protective device to prevent the occurrence of something the language seems unable to tolerate is thus another feature that unites Caddo and Seneca, even though the intolerable conditions are not identical in the two languages. One last resemblance that seems worth mentioning here is that which exists between Caddo -t- 'benefactive' and Seneca -at- 'reflexive' (or 'middle voice'). There is an obvious similarity in shape. Both elements, moreover, occur as the leftmost constituents of verb bases. In addition, both affect transitivity in some of their occurrences, although in opposite ways. The Caddo 'benefactive' sometimes transitivizes bases otherwise intransitive, while the Seneca element sometimes has the opposite effect.

2.1.4

Macro-Siouan

phonology

The phonological inventories of Caddoan languages appear superficially to be quite different from those of either Siouan or Iroquoian languages. In traditional phonemic terms, for example, Caddo has three vowels and nineteen consonants while Seneca has seven vowels and eight consonants. It is interesting and suggestive that the greatest differences between these various languages show up at the "systematic phonetic" level, while at the morphophonemic or "systematic phonemic" level, which undoubtedly reflects an earlier stage of the languages' history, they look much more alike. Finally, when one reconstructs phonological inventories for the three proto-languages, the similarities are even more striking. It is useful to line up these reconstructed inventories side by side in a way that will emphasize their common characteristics : Proto-Siouan l u e 0 a i Ρ w s h

a t y s ?

Proto-Northern-Iroquoian ι e 0 a

H

?

k r X

(ρ) w s h

Proto-Caddoan ι a

Q t y

η ?

k r

Ρ w s h

t y ?

There are controversial points in each of these reconstructions. The nasal consonants m and η have been omitted from the Proto-Siouan

53 inventory on the advice of Terrence S. Kaufman, who argues that these consonants in the modern Siouan languages can be explained as reflexes of other consonants or consonant clusters which were followed by nasalized vowels. The labial stop ρ is included in the Proto-NorthernIroquoian inventory on the basis of suggestions which have been made by Floyd G. Lounsbury and Paul M. Postal. The Proto-Caddoan list differs slightly from that published in Taylor (1963b). His c, kw, and ks have been omitted on the grounds that they may ultimately be explainable as clusters, and y has been added because it seems to be needed to explain developments in Caddo. Affricates, labiovelars, and résonants may be considered to be the most problematic aspects of Caddoan linguistic history at the moment. However these questions are resolved, the similarity between the three families will not be greatly affected. The point is that the phonological inventories of Siouan, Iroquoian, and Caddoan appear to converge as one traces them back in time. This kind of evidence for relationship may not be inconsequential when it is coupled with evidence of other kinds.

2.2

OTHER POSSIBLE REMOTE RELATIONSHIPS

The relationships which link the members of the Macro-Siouan stock to each other are not the only remote relationships involving these language families that have been suggested. Most important are the several relationships outside this stock that have been suggested for the Siouan family. There is, of course, the fact that Sapir included Siouan, along with Iroquoian and Caddoan, in his extensive and diverse Hokan-Siouan stock. More interesting, however, is Sapir's suggestion of a special closeness of relationship between Siouan and Yuchi. Yuchi is a linguistically isolated language originally spoken in Tennessee and the western part of the Carolinas, and presently still spoken by a few people in Oklahoma. The principal description of the language is Wagner (1934); there is a more recent but brief article (Wolff 1948), as well as an analyzed text (Wolff 1951). Further work is currently being performed by James M. Crawford and W. L. Ballard. Mary R. Haas has published a few lexical resemblances which support Sapir's hypothesis of a Siouan-Yuchi relationship (Haas 1951). Crawford (1971) points to resemblances between Yuchi and Ofo-Biloxi. Haas (1964) presents some evidence that Siouan is related not only to Yuchi, but also the Nadene languages (specifically Athapaskan and Tlingit). Haas (1951), as well as Haas (1952), also suggests a relationship between Siouan and the "Gulf languages" (the

54 Muskogean family plus Natchez, Tunica, Chitimacha, and Atakapa), based on reconstructions of the words for 'water' and 'land'. It may be noted that if Siouan is related to the Gulf languages and the latter in turn are related to Algonquian (Haas 1958), there is essentially but one superstock of languages east of the Rocky Mountains. A relationship as geographically far-flung as that suggested by Haas between Siouan and Nad'ene has been given some documentation by William W. Elmendorf, who has found some resemblances between Siouan (as well as Yuchi) and the Yukian languages of California (Yuki, Huchnom, Wappo) (Elmendorf 1963). He was led to look for such resemblances by an earlier suggestion of Paul Radin. Elmendorf (1964) presents further evidence for the Siouan-Yuchi-Yukian relationship. If all these suggestions are valid, they mean that Macro-Siouan is only part of a great linguistic stock which covered North America east of the Rockies, which extended with Nadene into the Northwest, Southwest, and all the way to the West Coast, and which had a further West Coast representative in Yukian, the latter perhaps providing a further link to Penutian, Hokan, or both. For the moment it would be well to remain aware of the speculativeness of these ideas.

A BRIEF LOOK AT THE CADDO LANGUAGE

3.1 3.1.0

CADDO PHONOLOGY

General remarks

As mentioned in the Introduction, Caddo belongs to a type of language that can be termed both polysynthetic and fusional. Its words (more correctly, a significant proportion of its verbs) are likely to contain a fairly large number of distinguishable parts (or morphemes), and these parts have been more or less radically fused together as the result of phonological change. In this section on phonology I will describe some of the processes which are responsible for the fusion. In the following section we will see some of the processes which underly the polysynthesis. Whether and to what extent it is desirable to account for the phonology of a language in terms of abstract underlying phonological shapes which are subjected to a series of phonological processes is at present an open question, perhaps ultimately resolvable through determinations of psychological relevance. The remarks I shall make here are at least valid within the domain of internal reconstruction, where the underlying forms are those inferred to have existed at some earlier stage of Caddo, and the processes applied to them are inferred changes which have taken place during the language's phonological history. Whether some or all of these forms and processes capture something that a present-day speaker of Caddo in some sense 'knows' about the language, I leave to the reader to decide according to his preferences. I have been able to identify at least 60 distinct phonological processes (or internally reconstructed phonological changes) in Caddo. I will not attempt to set forth even a significant proportion of them here, but will simply describe a few which stand out as particularly salient.

56 3.1.1

Surface and underlying phonological

inventories

The segments which are distinguishable phonetically include the following. There are only three vowels : i, u, and a. The first two have a pronunciation which is consistently rather low for high vowels ([i] and [u]). The low vowel a, furthermore, is raised in the direction of schwa in a closed syllable. This raising takes place not only in words like kat 'knife', which sounds much like English cut, but also and rather strikingly before y and w, as in 9ahay 'yes' and dahaw 'onion'. The final syllables of these words sound quite different from the diphthongs in English high or how, the pronunciation of aw, for example, being closer to that found in Canadian English in words like house. (Earlier observers of the language sometimes transcribed these diphthongs as e and o.) The largest class of consonants consists of the oral obstruents : Voiceless stops Voiced stops Glottalized stops Sibilants Affricates Glottalized affricates

Ρ b

t d t' s c c'

k k' s c c'

Other consonants include the laryngeal obstruents ? and h, the oral résonants w and y, and the nasal résonants m, η, and η. Vowels occur both short and long, and I will use a raised dot to indicate vowel length. Actually, this length is a property of the entire nucleus of a syllable, which may consist of a vowel plus a following resonant. In the latter case it is the entire sequence which is long, and I will place the raised dot after the resonant: háwt'uh 'wind', dankih 'six'. Consonants may be long as well, but I will represent consonant length as gemination. There are also significant pitch differences between syllables; the language has what might be regarded as a 'pitch accent'. In general, a syllable may have a pitch that is either high or low, and I will represent high pitch with an acute accent mark over the vowel. Low pitch will be left unmarked. Some syllables, however, have a sharply falling pitch, which I will represent with a circumflex accent. For reasons which will be seen below, the falling pitch occurs only in syllables whose nucleus is long. It would appear that at an earlier stage of the language the consonant inventory was smaller, consisting only of the stops p, t, and k, the

57 sibilant s, the laryngeal obstruents ? and h, and the three résonants w, y, and n. Whenever their origin can be reconstructed, it is clear that the glottalized consonants arose from a sequence of stop or affricate followed by

(1) (The symbol S stands for any stop or affricate.) The voiced stops, moreover, seem to be reflexes of the corresponding voiceless stops — that is, current b and d come from earlier ρ and t — except that the wordinitial d is often a reflex of earlier y :

(2) (3)

ρ, t —> b, d y-+d/#-

(The symbol # stands for word boundary.) The present voiceless ρ and t have various origins. In word-initial position they come from earlier w and η :

(4) w, η ->p, t/# — but they also arose in various ways from word-internal clusters. The palatal sibilant s and all the affricates (c, c, and their glottalized counterparts) arose through palatalization (see below). The nasals m and η are largely the products of the assimilation of « to a following consonant, but geminate mm resulted from the cluster nw. There seems to have been a dialect in which m replaced w, at least in some positions, and some words may have been borrowed from that dialect into the variety of Caddo described here. Although internal reconstruction suggests that the consonant inventory was once restricted to that described at the beginning of the last paragraph, it is not, I think, justifiable to attribute such a reduced inventory to synchronic underlying phonological shapes in all cases. There are many cases where the absence of alternation makes it impossible to retrieve an underlying ky?, let us say, as underlying a phonetic mm which is only one example of the widespread processes of assimilation and other kinds of reduction which operate on the clusters produced by syncope, as well as on clusters that would have existed in any case. Whether these processes were already operative prior to the second syncope is usually difficult to decide. Some of them appear to be persistent processes repeatedly applied whenever the necessary conditions arose. (Rule 1 probably expresses a process of this sort.) For most of them, however, it is only necessary to posit their application subsequent to the second syncope. I will simply list here some of those which are most frequently encountered. Rule 10 illustrates the fact that dentals are regularly assimilated in position of articulation to the following consonant in a cluster, often with attendant changes in the manner of the following consonant as well. We see something of the same sort in: (11) tw -»·pp

61 For example hat-wa-háh-nah 'they traded' appears phonetically as happaháhnah. If the second consonant was a stop, we find complete assimilation. Thus, for example : (12)

tk^kk

as in nas-a-na-batak-ah 'when it has gotten warm', which appears phonetically as nasámbakkah (after losing its third vowel by the first syncope and its fifth vowel by the second syncope). This example also shows the assimilation of « to a following labial, the process which is the principal source of m in the language, just as the parallel assimilation to a velar is the sole source of η : (13) n -> m, η / labial, velar It is interesting to observe the effect of the introduction of the incorporated noun root -kan- 'liquid' into the above word. Because the first syncope is excluded from nas-a-kan-na-batak-ah 'when the liquid has gotten warm', the second syncope applies tò a different vowel (the fifth in this form), producing nasakannawtakah, whose final portion is quite different from that in the shorter word cited above. The word just cited shows a change of b to w following the syncope. Strictly speaking, this process is not one of cluster reduction but rather an example of a class of processes which involve some kind of 'weakening' in syllable-final position : (14) b -> w I when syllable-final It is, then, the fact that syncope places the b in syllable-final position that brings about its change to w. Something similar happens to affricates, which 'weaken' to spirants : (15) c, ë ->s, s I when syllable-final For example, one may observe the fate of bak-yi-bahw-nah 'he heard', where ky first becomes c by the second palatalization, the second vowel is dropped by the second syncope, and the £, now in syllable-final position, becomes s. The result is basbáwnah (which includes in its penultimate syllable a phenomenon yet to be described). In this same category belongs the weakening of k to h : (16) k

h Iwhen syllable-final, except before another k

For example, ci-ki-duk-ah Ί bit' appears phonetically, after syncope of

62 the second vowel and the application of rule 16, as cihdukah. The development of high pitch in originally low-pitched vowels takes place in several ways. We have already seen that compensatory pitch raising is one aspect of the first syncope (rule 8). Compensatory pitch raising is also associated with the loss of h before two consonants : (17) V h C C - V C C (This process fails to apply when CC consists of a glottal stop followed by a resonant.) An example was given in the preceding paragraph, where we saw bak-yi-bahw-nah realized phonetically as basbáwnah. It is also the case that an originally low-pitched vowel copies the high pitch of a following vowel if it is separated from the latter by a single resonant : (18) V R V - V R V For example, ni-wàhd-ah 'he brought' appears phonetically as tíwáhdah, with high pitch on the first vowel as well as the second. The addition of length to originally short syllables also takes place in several ways. An open penultimate syllable which is high-pitched automatically becomes long, as may be expressed in the following way: (19) V(R)CV -> V(R) CV / if the second of the two vowels is the last in the word We saw an example of this in the words háwt'uh 'wind' and dàn'kih 'six' given earlier. It can be seen after a simple vowel nucleus in wá 9ah 'wasp' or kadi'sah 'he washed' from ka-dis-ah. The résonants y and w, when preceded by the corresponding vowels i or u, become simply a lengthening of the vowel : (20) iy, uw -> r, w For example, ku-bak-yi-bahw-nah 'he heard me' becomes kuwcibáwnah as a result of rule 14 and various others set forth above, but then rule 20 leads to ku ëibâwnah. It is also the case that in any syllable prior to the penult a syllable-final glottal stop is converted into lengthening of the syllable : (21) V(R)

V(R)'/in prepenultimate syllable

(This rule does not apply before a glottalized consonant.) For example, ni-?a-dih-'?a7 'he will take' loses its second vowel, placing the ? in

63 syllable-final position. The process just described then yields the phonetic form ti'dih 9 It is in association with rule 21 that falling pitched syllables arise, specifically when the vowel in the environment described in rule 21 is high pitched. We can say that before rule 21 applies, such a vowel takes on a falling pitch : (22) V(R)

V(R) ? /in prepenultimate syllable

whereupon rule 21 replaces the ? with length. As an example we can look at what happens to yah?-?a-dih-hah 'you go'. High pitch is introduced into the first vowel by rule 17, which also deletes the h. The following pair of glottal stops is simplified to one by a process which degeminates laryngeal obstruents, and which also applies to the hh cluster later in the word : (23)

AA->

h

The second vowel is then syncopated, and the initial syllable of the word is now yd?, making it subject to rule 22. The resulting ya7 is further subject to rule 21, to yield yâ'. Taking rule 3 also into account, the final phonetic shape is dâ'dihah. An example in which the falling pitched syllable contains a resonant can be derived from the underlying form yah?-yi-bahw-nah 'you saw'. Again rule 17 deletes the first h and adds high pitch to the first vowel. The second vowel is then syncopated, giving an example of syncope after a sequence of glottal stop plus resonant (as stated in parentheses immediately after rule 9). At this point the sequence 7y is metathesized to yas regularly occurs to glottal stop plus resonant in syllable final position : (24) ?R - R ?/when syllable-final The first syllable of this word is now ydy If we again apply rules 22 and 21, as well as 3, we arrive at the final phonetic shape dâybàwnah. These last examples show the kind of complexity that Caddo words often contain. The processes that have been described are all frequently operative in the language, as are others that have not been touched on. As various grammatical features are discussed in the next section, these processes will be frequently illustrated and a few additional ones will be introduced as necessary. Pending a more complete phonological description of the language, what is given here will provide the reader with a

64 reasonably clear impression of what Caddo phonology is like. One last word on phonology is in order. There is in the language at present a distinction between certain 'fast speech' and 'slow speech' forms, the former being those normally used in conversation. Particularly evident in fast speech is the dropping of h and w between vowels. Thus, the usual Caddo greeting is in slow speech kúha?ahat (based on the adjective 'good' with the prefix kú- 'just' ; literally 'it's just good'). Usually this word is pronounced kúa ?a't, where the two h 's have been deleted and the resulting sequence aa is equivalent to a'. The loss of intervocalic w in fast speech can be illustrated on the basis of the underlying form ci-wiht-^a-dih-hah 'we (dual) go'. Processes already described, plus one that deglottalizes a glottalized stop before another consonant : (25) S'C^SC

(here

t'd-^td)

lead to the slow speech phonetic form cíwítdihah. The corresponding fast speech form is ci'tdihah. Thus, we may say that the following rule applies in fast speech : (26) A, w -> Ç) /V—V It may also be mentioned that rules 21 and stop with length and change high pitch to speech to any prepenultimate syllable of a word. Thus slow speech di9 ci9 da ?ah 'dog speech di'ci' da9ah.

22, which replace a glottal falling pitch, apply in fast phrase, and not just of a be' is pronounced in fast

3.2 C A D D O G R A M M A R 3.2.0

General remarks

As discussed in the Introduction and mentioned again at the beginning of the last section, the remarks which follow will focus on certain factors which are responsible for the polysynthetic nature of the Caddo verb. Not unlike a number of other American Indian languages, Caddo achieves its polysynthesis in part through the absorption into the verb of various elements which lie semantically outside it. More specifically, various of the parts which one finds included in the surface-structure verb have their semantic origin in one or more of the nouns associated with

65 that verb. The discussion here will begin by focusing on three distinguishable kinds of nominal elements which are thus attracted into the verb. The first are features of person and case (with which a sentence feature of 'unreality' may be combined); the second are features of number; and the third is the movement of the entire lexical content of certain nouns into the verb (that is, noun incorporation). After that I will describe various features which originate semantically as modifiers or inflections of the verb itself, to show how they also add to the total complexity of the verb in surface structure. Those features to be treated here will include the following: a benefactive derivational element; elements which specify whether the verbal meaning is accomplished in a sitting, standing, or lying posture; features of tense and aspect; lexical gap questions; nominalization; the formation of locative and temporal adverbial clauses; and features which convey the speaker's understanding of the sentence's validity. The theoretical assumptions on which this account is based are set forth in Chafe (1970a, 1970b).

3.2.1

Person, case, and reality

If any noun accompanying a verb contains either the feature 'first person' or the feature 'second person', that feature always moves into the surface structure verb, bringing with it at the same time an indication that the noun from which it came was either the 'subject' or 'object' of the verb. The semantic origins of these subject and object roles are complex and will not be treated here. (For example, the language is at least in part of the 'ergative' type, so that surface subjects and objects bear some correlation to semantic agents and patients respectively.) Suffice it to say that surface structure verbs in this language exhibit prefixes characterizable as 'first person subject', 'first person object', 'second person subject', and 'second person object'. In addition, it is possible for a third person noun to be characterized as 'indefinite' — that is, to refer to people in general rather than to any specific person or thing. (Such a noun might be translated 'one', French 'on', or German 'man'.) Prefixes reflecting such an 'indefinite subject' or 'indefinite object' are also found in surface structure verbs. Aside from this indefinite type, however, the person and case of a third person subject or object noun is not represented by any surface verbal prefix. The underlying phonological shapes of the prefixes just described are as follows :

66 cikuyah siyiyu-

first person subject first person object second person subject second person object indefinite subject indefinite object

Examples of their use are given in the following paradigm of the verb -yi-bahw- 'see' with the past tense suffix -nah : ci-yi-bahw-nah —• cí'báwnah ku-yi-bahw-nah kúybáwnah yah ?-yi-bahw-nah -> dâybàwnah si-yi-bahw-nah -> svbdwnah yi-yi-bahw-nah -> dí'báwnah yu-y i-bahw-nah —• dúybáwnah

Ί saw' 'saw me' 'you saw' 'saw you' 'one saw' 'saw one'

A particularly interesting feature of Caddo is the fact that these prefixes appear in an entirely different shape in questions, negations, conditions, and other sentences of a hypothetical nature. The term 'unreal' may be used for this second set of prefixes, which are thus in contrast to the 'real' prefixes described above. The unreal prefixes may be considered to result from the addition of a feature of unreality or hypotheticalness which fuses with the prefixes listed above. The resulting shapes are: t 'abasah sa7a-

first person subject unreal first person object unreal second person subject unreal second person object unreal indefinite subject unreal 7 7 a aindefinite object unreal

When these unreal prefixes are simply substituted for the corresponding real prefixes, the result is a yes-no question : t'a-yi-bahw-nah -> t'áybáwnah 'did I see?' ba-yi-bahw-nah ->• báybáwnah 'did — see me?' sah 7-y i-bahw-nah —> sây'bàwnah 'did you see?' sa 7a-yi-bahw-nah —> sa ?áybáwnah 'did — see you?' ? a-yi-bahw-nah —> ?áybáwnah 'did one see?' ? a ">'a-yi-bahw-nah —> ?a ?áybáwnah 'did — see one?'

67 There is also a prefix sa-, which I believe is best interpreted as representing the 'unreal' feature in isolation. It occurs in those cases where no person-case prefix is present ; that is, when there is semantically no first person, second person, or indefinite subject or object noun accompanying the verb, but only a third person subject and/or object which is not indefinite. We find, therefore: sa-yi-bahw-nah -> sáybáwnah

'did (he/she/it) see?', or 'did—see (him/her/it)?'

Other uses of these unreal prefixes appear when they are preceded by certain other prefixes such as kú- 'negative', hi- 'conditional', or nas'whenever' or 'when in the future' : kú-t'a-yi-bahw-nah hí-t'a-yi-bahw nas-t'a-yi-bahw

kút'áybáwnah hít'áybah nast'áybah

Ί didn't see' 'if I see' 'when I see'

The pure unreal prefix which appears in word initial position as sa- has the shape -ya- when preceded by kú or hi-, and the shape -a- when preceded by nas- : kú-ya-yi-bahw-nah

kúyáybáwnah

hí-ya-yi-bahw -*

híyáybah

nas-a-yi-bahw -*

nasâybah

'(he/she/it) didn't see', or 'didn't see (him/her/it)' 'if (he/she/it) sees', or 'if — sees (him/her/it)' 'when (he/she/it) sees', or 'when — sees (him/her/it)'

It is possible for a verb to be accompanied by both a first person and a second person noun. Both features will then be absorbed into the surface structure verb, to yield the following composite prefixes : t'afirst person subject, second person object yahkusecond person subject, first person object Examples are : t 'a-yi-bahw-nah -* t 'áybáwnah yahku-yi-bahw-nah —• dahkúybáwnah

Ί saw you' 'you saw me'

The shape of the first of these is no different when the unreal feature

68 is added, but the second becomes sahku- when unreal : t'a-yi-bahw-nah —• taybàwnah sahku-yi-bahw-nah —• sahkúybáwnah

3.2.2

'did I see you?' 'did you see me?'

Number

Under certain circumstances the number specification of one or more accompanying nouns is also copied into the surface structure verb. The manner in which this happens is somewhat complicated, and may be described in terms of the following three processes, each of which applies whenever it can : (a) If an object noun is nonsingular (either dual or plural), this fact is reflected in the presence of a 'nonsingular object' prefix in the surface verb. Its shape is haka- or na-, depending on the verb. (b) If either a subject or an object noun (or both) is dual, this fact is reflected in the presence of a 'dual' prefix in the surface verb. Its shape is wiht-. (c) If either a subject or an object noun (or both) is plural, and if that noun is also animate, this fact is reflected in the presence of a 'plural animate' marker in the surface verb. Its shape is -wa-. It occurs between the two parts of the verb stem — the preverb and the verb proper (see p. 45).

The relative order of all these parts of a surface structure verb is as follows : 1 person case

2 dual

3 nonsingular object

4 preverb

5 plural animate

6 verb proper

The following are examples with the same verb stem -yi-bahw- 'see' and the past tense suffix -nah. It is assumed that the accompanying nouns are third person animate. It may be noted that the fourth and fifth, sixth and eighth, and seventh and ninth words have the same surface form:

69 Number of Subject sg·

Number of Object sg·

du.

sg-

pl.

sg·

sg.

du.

du.

du.

pl.

du.

sg-

pi-

du.

pi.

pl.

pl.

yi-bahw-nah —• dibdwnah '(he) saw (him)' wiht-yi-bahw-nah -> písbáwnah 'they (du.) saw (him)' yi-wa-bahw-nah —> diwabáwnah 'they (pi.) saw (him)' wiht-haka-yi-bahw-nah —• píkkáybáwnah '(he) saw them (du.)' wiht-haka-yi-bahw-nah —> píkkáybáwnah 'they (du.) saw them (du.)' wiht-haka-yi-wa-bahw-nah -> pikkáywabáwnah 'they (pi.) saw them (du.)' haka-yi-wa-bahw-nah —> hakáywabáwnah '(he) saw them (pi.)' wiht-haka-yi-wa-bahw-nah —• pikkáywabáwnah 'they (du.) saw them (pi.)' haka-yi-wa-bahw-nah -* hakáywabáwnah 'they (pl.) saw them (pl.)'

Any such word may also include a person-case prefix, placed before the number prefixes, to reflect the person and case of one or more of the accompanying nouns. For example, with the prefix t'a- 'first person subject, second person object' : Number of Subject

Number of Object

sg-

sg-

du.

sg·

pl.

sg-

sg-

du.

t 'a-yi-bahw-nah —> Ί saw you (sg.)' t'a-wiht-yi-bahw-nah -» 'we (du.) saw you (sg.)' t 'a-yi-wa-bahw-nah —> 'we (pi.) saw you (sg.)' t 'a-wiht-haka-yi-bahw-nah etc.

t 'áybáwnah t 'áwisbáwnah taywabáwnah t 'áwikkáybáwnah

Caddo also distinguishes between 'exclusive' and 'inclusive' first persons. The forms with ci- carry the exclusive meaning when they are nonsingular. For example :

70 ci-wiht-yi-bahw-nah -» cíwísbáwnah ci-yi-wa-bahw-nah -* cí'wabáwnah

'we (excl. du.) saw' 'we (excl. pi.) saw'

Interestingly enough, the inclusive forms contain the same prefixes which, when they are singular, carry the 'indefinite' meaning. What is indefinite in the singular is inclusive in the nonsingular : yi-wiht-yi-bahw-nah -> díwísbáwnah yi-yi-wa-bahw-nah —• dí'wabáwnah

3.2.3

'we (incl. du.) saw' 'we (incl. pi.) saw'

Incorporation

Caddo nouns fall into two major classes with respect to their surface structure behavior : those that incorporate and those that do not ; that is, some nouns move postsemantically into the surface structure verb while others do not. On this basis we may speak of 'incorporating' and 'nonincorporating' nouns. The distinction is to a large extent arbitrary from a semantic point of view; for the most part one cannot predict from the meaning of a noun whether it will belong to the incorporating or non-incorporating type. However, nearly all body part nouns are incorporating. Most but not all incorporating nouns are capable of occurring in isolation — that is, in unincorporated position — in which case they are accompanied by the general noun suffix -?uh. For example, the noun -c'ah- 'eye' is found in the word c'ah?uh 'eye' as well as incorporated within a verb. If a speaker of the language is asked the word for 'eye', however, he is just as likely to employ incorporation by saying kah ?ic'áy 9ah, which is really a nominalization of the phrase 'be an eye' : literally 'that which is an eye'. The underlying form of this word is kak-?i-c'ah-ya?ah, where kak- is the 'that which' nominalizer, - ? i- is an empty morpheme that occurs before many noun roots, and the final ya7ah is the verb 'be'. Similarly, 'bag', whose underlying shape is -bäht-, may be given as bäht 'uh or kahbâhca ?ah, the latter being literally 'that which is a bag'. For some nouns, however, the form with -?uh does not occur at all. For example, the only way for a Caddo to answer the question How do you say 'house'? is to give the word kahnisdy?ah, literally 'that which is a house', which contains the incorporated noun -nisah- 'house'. The difference in the surface treatment of incorporating and nonincorporating nouns can be seen in pairs of sentences like the following,

71

the first four of which contain the verb stem -ka- ?ni'buy' (with high pitch assigned to whatever vowel precedes the portion and the past tense suffix which in this case has the shape -ah. The symbol # is used for word boundary : nisah-ka'- ?ni ?-ah tisahkâni ?ah 'He bought a house.' ka ?ás # ka-9ni7-ah -»· ka 9ás kâ'nPah 'He bought an apple.' ci-nisah-ka- ?ni ?-ah —• ci'sahkâ'ni ?ah Ί bought a house.' ka?ás # ci-ka'-?ni7-ah —> ka?às cikâ'ni?ah Ί bought an apple.' di' # nisah-ya ?ah —• di' tisáy ?ah 'This is a house.' di' # ka ?ás # ya ?ah -> di' ka ?ás da ?ah 'This is an apple.' (In the third word the η of -nisah- becomes y before s after syncope of the second vowel, and then subsequently becomes length after the vowel i.) An interesting feature of incorporation in Caddo results from the fact that certain nouns which do not themselves incorporate have the property of being included semantically within other, 'classificatory' nouns (Chafe 1970b: 116-8) which do incorporate. As examples we may take the nouns ?iniku? 'church' and cúcu7 'milk'. A church is conceived of as a kind of house (or building), and milk as a kind of liquid. The nouns 'church' and 'milk' are not incorporating, but the nouns 'house' and 'liquid' (phonologically -kan-) are. We then find incorporation of the larger, including concept, along with separate mention of the smaller, included one: iniku7 # nisah-ka'-?ni7-ah —• iniku' tisahkâ'ni?ah 'He bought a church.' cúcu ? # kan-ka- ?ni7-ah cú'cu' kœ^kâ'ni ?ah 'He bought milk.' di' # iniku ? # nisah-ya ?ah —• di' iniku' tisáy ?ah 'This is a church.' di' —• cúcu ? # kan-ya ?ah -> di' cú'cu' kan ?nah 'This is milk.'

72 (In the last word y assimilates to the preceding n, the second vowel is syncopated, and the glottal stop is metathesized into the middle of the resonant cluster.) It can be seen that incorporated nouns of the sort just described might be thought of as 'classifiers' within the verb, expressing a larger semantic unit within which the specific object noun is included. In some cases such an incorporated noun has taken on a specialized classificatory function distinct from the meaning it has when it is not used in this way. A clear example is -c'ah-, which carries the meaning 'eye', as we saw above. It is used in a classificatory way, not for objects which are literally 'kinds of eyes' (whatever they would be), but for small round objects. Thus we find it incorporated in the verbs of sentences like the following: kassi ? # hák- ?i-c 'ah-s-swí ?-sa ? -> kassi? háh ?ic 'ásswí ?sa ? 'She's stringing beads.' ka ?ás # hák- ?i-c'ah- Ή ?-sa? ->ka ?ás háh ?ic'ah ?í ?sa ? 'Plums are growing.' (The 'progressive' aspect evident in these two verbs has a complex surface representation consisting of the prefix hák- and the suffix -sa

3.2.4

Benefactive

Some verbs which are not (like 'give') inherently benefactive — which do not in themselves require the accompaniment of a beneficiary noun—may be made so derivatively. This kind of derivation, which occurs with considerable frequency, is reflected in surface structure through the appearance of a benefactive (or dative) prefix which occurs at the very beginning of the verb stem. The underlying shape of this prefix varies according to the stem which follows it. Most commonly it is either -t-, -'n- (that is, η with high pitch on the preceding vowel), -'ni-, or simply -«-. Another peculiarity is that a third person beneficiary noun, unlike either a third person agent or patient, has its presence reflected in the surface verb, although only if no first or second person prefix is present. The underlying shape of this 'third person beneficiary' prefix is nu-. Various other irregularities are associated with these benefactive constructions; for example, the 'first person subject, second person object' prefix has the shape t 'u- rather than t 'a- when followed by the benefactive prefix. In the following examples there is first a word without the

73 benefactive, then several words containing it. In the first examples it has the shape -t- : háy-yúh-cah —> háyyúhcah '(he) is going to tell' nu-t-háy-yúh-cah —> tutháyyúhcah '(he) is going to tell to him' ci-t-háy-yúh-cah —• citháyyúhcah Ί am going to tell to (someone)' t 'u-t-háy-yúh-cah -* t 'utháyyúhcah Ί am going to tell to you' In the following example the benefactive prefix has the shape - n - : ?

i-kiyu?-ah -> 'Hhyu ?ah '(he) killed' ku'-n-kiyu ?-ah kúncí'9ah '(he) killed for me' (The empty prefix Ή- discussed on p. 51 can be observed in the first of these words. In the second the benefactive prefix changes a following k to c.) In the next example the benefactive has the shape -'nt- : ?

i- ?áh-nah —• ?i ?dhnah '(he) ate' ku'-nt- ?áh-nah -+ kúnt'áhnah '(he) ate for me' In the following the shape is -«-: ?

i-bí ?n-cah —> ?ibin ?cah '(he) is going to wipe' ku-n-bi 9n-cah —• kumbin ?cah '(he) is going to wipe for me' The benefactive construction is also used to convey possession, usually with the verb stem -ya ?ah 'be' or the verb stem - ?a? 'be present, exist' : ku-t-ya 9ah -* kuc'ah 'is to me', 'is mine' hák-ku'-nt-7a 7 —• hákkúnt 'a ? 'is present to me', 'is mine'

74 hák-ku-t-cakah- 9a 9 -* hákkutcakah 9a 9 Ά belt is present to me', Ί have a belt.'

3.2.5

Sitting, standing, and lying

A great deal could be said about the various semantic elements which lead to increments within the surface verb stem, beyond the phenomena of incorporation and benefactiveness just described. I will go no further here, however, than to illustrate something that was mentioned earlier on p. 47: the fact that verbal notions are often conceived of as occurring or existing while the person involved in them is in a sitting, standing, or lying posture. Such notions are conveyed through the addition of steminitial elements which have the shapes - 9awis- 'sitting', - 9anikis- 'standing', and -9ini- 'lying'. Thus, while one may say simply: hák- 9a-nát-sa 9 —> háh 9ánássa 9 '(He) is cold.' one may also say : hák-7aw is- 9a-nát-sa 9 —> háh7aw is 9ndssa 9 '(He) is cold while sitting.' hák- 9anikis- 9a-nát-sa 9 —• háh 9ánkis 9nássa ? '(He) is cold while standing.' hák- '•'ini- ^a-nát-sa? -* háh9 ini' nassa? '(He) is cold while lying.' To say simply that someone is sitting, standing, or lying it is necessary to join these elements to either -ya ?ah 'be' or - ?a? 'be present, exist' : ?

awis-ya 9ah —• ?dwsa ?ah '(He) is sitting.' ? anikis-ya 9ah -> ? háh 9ca]kisa 9 '(He) is standing.'

75 hák-?ini-9a ? —• '(He) is lying.'

háh ?in ?a?

(In the fourth and fifth words it can be seen that s? is simplified to J.)

3.2.6

Tense and aspect

Semantically, a verb is nearly always inflected in terms of some particular tense or aspect. (I am using these terms together as a complex label for the class of inflections I have in mind, and will not try to distinguish some as tenses and others as aspects.) Among the most commonly occurring inflections of this type are the following, given with their phonological representations. It can be seen that most are suffixes, though one is a prefix and one a prefix-suffix combination : -nah or -ah hit-wa ? or - ?a 7 Sah -hah hák-... -sa?

'past I' 'past II' 'future I' 'future II' 'iterative' 'progressive'

The difference in meaning between past I and past II is not entirely clear. Both are used in the communication of situations and events that occurred in the past. Sometimes past I is used with a perfective or 'current relevance' meaning, appropriately translated by English have with the past participle, but that seems not always to be the case. Past II, however, never has the perfective meaning. The choice between -nah and -ah for 'past Γ is in part phonologically based, with -ah occurring consistently after oral obstruents. After other consonants, however, there seems to be little phonological predictability, the choice being dictated by the particular morpheme that precedes. We had examples of past I with the verb stems meaning 'see' and 'buy' above. The following show the 'past II' prefix: hit-ci-yi-bahw -» hitcvbah hít-ci-ka'- ?ni? —> hítciká 7ni ? hít-yi-bahw -> hítdibah hít-ka-7ni? —• híkká ?ni?

Ί saw' Ί bought' '(he) saw' '(he) bought'

76 In the third word it may be noted that the change of underlying y to d, previously described as occurring in word-initial position, takes place even though there is a preceding prefix. This prefix belongs to a class which might be termed 'proclitic'. Proclitic prefixes behave, in respect to certain but not all phonological processes, as if they were separated from what follows by a word boundary. We might indicate such a semi-word boundary with a plus sign. Thus, the third word above has an underlying form which can be written hit+yi-bahw, and the change of y to d can be said to take place after + as well as in word-initial position. That + is not a complete word boundary can be seen from the assimilation of the / to the following k in the fourth word. It is likely that proclitic prefixes like hit- were once separate particles, which have now become partially attached to the following word. The historical process by which word-initial y became d would thus have run its course by the time of this attachment, but the assimilation of t to k would have applied subsequently. (With respect to the phonology of the above examples, it should also be noted that the change of ά? to cikâ'nPèah

Ί Ί Ί Ί

will see' will buy' am going to see' am going to buy'

The iterative and progressive inflections are both used in 'present tense' contexts. The iterative is very often found in the generic use, although it may also convey a particular, repetitive event. The progressive seems to function in much the same way as the English progressive {be with the present participle), expressing a limited duration which precedes and follows the moment of utterance. Examples with the iterative are : ka-dis-hah —• kadishah kak- ?á ?-hah -> kak 'â ?hah

'(he) washes' '(he) cries'

77

The same verbs in the progressive are : hák-ka-dís-sa ? —• hákkadíssa 7 hák-kak- ?ά 7-sa -> hákkak a ?sa ?

'(he) is washing' '(he) is crying'

(The same translations are sometimes given for the iterative.) The iterative inflection may occur in conjunction with past II, in which case either a repeated or durative action in the past is indicated : hít-ka-dís-hah —• híkkadíshah hit-kak- 7ά?-hah -* híkkak a?hah

'(he) was washing' '(he) was crying'

It may be that whatever distinction exists between iterative and progressive is neutralized in these past-tense forms. The only surface structure verbs which do not occur with some overt tense-aspect inflection are imperatives, consisting of the plain verb stem with a second person subject prefix: yah 9-yi-bahw —• dáy ?bah yah ?-bak-yi-bahw —• dâw öibah

'look ! ' 'listen ! '

(In the latter word the stem -yi-bahw- 'see' is preceded by the incorporated noun -bak- 'voice, word'. In other words, 'listen, hear' is an idiom whose literalization is 'see the voice'. The inference is thus that seeing was at one time conceived of as the unmarked kind of perception.)

3.2.7

Lexical gap questions

We have already seen that yes-no questions are formed through the addition of the 'unreal' feature to the verb, where it appears in surface structure fused with the person-case prefixes. Lexical gap ('WH') questions are reflected in surface structure in the occurrence of various prefixes which are placed ahead of the person-case prefix. These lexical gap prefixes also may combine with past II to yield the shapes which are listed in the second column below : Non-past witdikat-

Past wi dikit-

'who?' 'what?'

78 kwítsit-

kwi'Htsi Ht-

'where?' 'when?'

The use of these prefixes is illustrated in the following, based on the verb stem -náy-?áw- 'sing' and the already familiar -yi-bahw- 'see'. (In various of these words it will be noted that an underlying high pitched vowel receives low pitch in a final syllable.) wit + náy- ?áw- ?a? -> wíttáy ?áw ?a ? wit+yah ?-yi-bahw -* widdy '•'bah wí ?ít+náy- ?áw -> wi Httây 7aw wíHt+yah 7-yi-bahw -> wPidáy ?bah

'Who will sing?' 'Whom do you see?' 'Who sang?' 'Whom did you see?'

dikat+náy- ?áw-> dikattáy ?áw ya ? ? dikat+yah -yi-bahw —>· dikadáy ?bah ? dikit + náy- áw dikittáy ?aw dikit+yah ?-yi-bahw -y dikidáy ?bah

'What 'What 'What 'What

kwít+náy- ?áwkwittáy ?áw?a? 9 kwít+yah -yi-bahw -> kwidáy ?bah kwi9 it+náy- ?áw —> kwi ?ittáy ?aw kwi?ít+yah ?-yi-bahw kwi?idáy ?bah

'Where 'Where 'Where 'Where

will (he) sing?' do you see (it)? did (he) sing? did you see (it)?'

sit + náy- ?áw- ?a? -» sittáy ?áw ?a? sit+yah ?-yi-bahw -> sidáy 7bah sí ?ít+náy- ?áw -* sí ittáy ?aw sí ?ít+yah ?-yi-bahw —• sí ?ídáy ?bah

'When 'When 'When 'When

will (he) sing? do you see (it)?' did (he) sing?' did you see (it)?'

will (he) sing?' do you see?' did (he) sing?' did you see?'

It may be noted that these prefixes are of the proclitic type, and hence separated from what follows by a + boundary. It can be seen that following η and y become t and d respectively, as in word-initial position. In addition, the final t of each prefix is lost before a following yah The same happens before the indefinite prefixes yi- and yu-, but not before any y that belongs to a stem. In this respect the initial y of the second person and indefinite prefixes is phonologically different from other instances of y.

3.2.8

Nominalization

There are various ways in which clauses may achieve a subordinate position—may be 'embedded'—within a larger sentence. One such

79 device is nominalization, of which I will describe two types. One of them we encountered earlier in the section on noun incorporation. We saw how a nominalized clause translatable as 'that which is' appears in surface structure with the prefix kak- attached to the verb. The examples given were : kak- H-ë'ah-ya ?ah -> kah ?ië'ày ?ah 'eye', literally 'that which is an eye' kah-báht-ya 9ah —> kahbáhca ?ah 'bag', literally 'that which is a bag' This kind of structure is found not only with the verb 'be' : kak-haka-wa- ?áh -> kahakáw 9ah 'hawk', literally 'that which eats things' (This is one of the exceptional words in which the vowel of the animate plural morpheme -wa- undergoes syncope.) The nominalized clause may modify either the pronominal subject or object of its verb. Furthermore, it may be used as an infinitive or gerund. There are, then, a variety of translations for a word like : kak-náy- ?áw -> kahnáy ?aw 'the one who sings' 'that which (he) sings' 'to sing' 'singing' or, when the subject is made explicit in the surface structure form : kak-ci-náy- 9áw —• kahcínáy 7aw 'for me to sing', 'that I sing' 'that which I sing' This nominalization may occur together with past II, in which case the form of the surface prefix is kik- : kík-náy- ?âw -» kíhnáy9aw 'the one who sang' 'that which he sang'

80 The other type of nominalization I will mention here has an instrumental function and is typically translated 'that with which' or 'that which is used for'. It appears in surface structure as the verb prefix nak-: nak-yi-ka-dis —• nakikadis 'soap', literally 'that with which one washes' nak-yi-s-swi? nakisswi? 'that with which one strings (beads)', referring to a needle, thread, or the like

3.2.9

Adverbial clauses

Other subordinate clauses function as sentence adverbials. Again, the kind of subordination is indicated by a surface verbal prefix. For example, a locative adverbial clause may be reflected in the prefix kúk-, translatable as 'where' : kúk- 9i-c'ah- ?a ? -> kúh ?i¿'ah ?a? 'where (his) eye is', 'in (his) eye' kúk-ci-yi-bahwkúhcí'báw7a? 'where I will see' When such a clause also contains past II, the prefix has the shape kwik-: kwík-ci-yi-bahw -> kwíhcí'bah 'where I saw' Adverbial clauses of time may be reflected in the prefix húk- 'while'. Usually such a clause also contains the progressive inflection, in which case húk- replaces the progressive prefix hák- but is still accompanied by the suffix -sa húk-ci-yi-bahw-sa -* húhcí'báwsa ? 'while I was looking' húk-náy- ?dw-sa ? -> húhrny ?áwsa? 'while he was singing' Nondurative temporal subordination may be reflected in the prefix nik'when' :

81 ník-ci-yi-bahw —• níhcvbah 'when I saw' We saw earlier that such temporal clauses in future time appear as the prefix nas- together with the unreal feature which is fused with the person-case prefix : nas-t'a-yi-bahw —• nast'áybah 'when I see'

3.2.10

Validity

The speaker's view of the validity of what he says may also be present as a semantic inflection, appearing again as a prefix on the surface structure verb. For example, if the speaker is repeating knowledge that he has received through language from someone else rather than through his own perception, he may make this explicit through the use of the prefix kán-: kán-yi-bahw -* kánnibah '(he) is said to see' kán-náy- ?ά\ν -> kánnáy9aw 'allegedly (he) sings / is singing' When combined with past II the form is kin- : kín-náy- 9áw —> kínnáy ?aw 'allegedly he sang' The speaker's assessment of the degree of probability by which the information is true is expressed through prefixes such as ták- 'possibly, might' and túk- 'probably' : ták-ci-yi-bahw- 9a? -> táhcrbáw ?a? Ί might see', 'possibly I will see' ? túk-ci-yi-bahwtúhcí'báw ?a ? 'probably I will see' A contrary assessment — that something is improbable — is expressed by the prefix wás- together with the unreal feature :

82 wás-í'a-yi-bahw -* wást'áybah 'I'm not likely to see' wás-sa-náy- ?ά\ν —• wásánáy ?aw '(he) is not likely to sing' This construction is often used also to express surprise at something that has turned out to be true after all : wás-ba-?a-sa-yik- ?awi-hah —